diff --git a/tests/models/aimv2/test_modeling_aimv2.py b/tests/models/aimv2/test_modeling_aimv2.py
index 085e307512fe..81e7e443eba1 100644
--- a/tests/models/aimv2/test_modeling_aimv2.py
+++ b/tests/models/aimv2/test_modeling_aimv2.py
@@ -18,7 +18,6 @@
 import unittest
 
 import numpy as np
-import requests
 from parameterized import parameterized
 
 from transformers import Aimv2Config, Aimv2TextConfig, Aimv2VisionConfig
@@ -44,6 +43,7 @@
     random_attention_mask,
 )
 from ...test_pipeline_mixin import PipelineTesterMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -58,8 +58,6 @@
 
 
 if is_vision_available():
-    from PIL import Image
-
     from transformers import AutoImageProcessor, AutoProcessor
 
 
@@ -479,7 +477,9 @@ def test_inference(self):
         model = Aimv2Model.from_pretrained(model_name, device_map=torch_device)
         processor = AutoProcessor.from_pretrained(model_name)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
         inputs = processor(
             text=["a photo of a cat", "a photo of a dog"], images=image, padding=True, return_tensors="pt"
         ).to(model.device)
@@ -513,7 +513,9 @@ def test_inference(self):
         model = Aimv2VisionModel.from_pretrained(model_name, device_map=torch_device)
         processor = AutoImageProcessor.from_pretrained(model_name)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
         inputs = processor(image, return_tensors="pt").to(model.device)
 
         with torch.no_grad():
@@ -544,7 +546,9 @@ def test_inference_for_native_resolution(self):
         model = Aimv2VisionModel.from_pretrained(model_name, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(model_name)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
         inputs = processor(image, return_tensors="pt").to(model.device)
 
         with torch.no_grad():
diff --git a/tests/models/aria/test_modeling_aria.py b/tests/models/aria/test_modeling_aria.py
index 9813f79ba95b..dbd9676d6aec 100644
--- a/tests/models/aria/test_modeling_aria.py
+++ b/tests/models/aria/test_modeling_aria.py
@@ -44,6 +44,7 @@
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -321,7 +322,9 @@ def test_small_model_integration_test_llama_batched(self):
             "USER: <|img|>\nWhat is this? ASSISTANT:",
         ]
         image1 = Image.open(requests.get(IMAGE_OF_VIEW_URL, stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=[image1, image2], text=prompts, return_tensors="pt", padding=True).to(
             model.device, model.dtype
@@ -359,7 +362,9 @@ def test_small_model_integration_test_batch(self):
             "USER: <|img|>\nWhat is this?\nASSISTANT:",
         ]
         image1 = Image.open(requests.get(IMAGE_OF_VIEW_URL, stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = self.processor(images=[image1, image2], text=prompts, return_tensors="pt", padding=True).to(
             model.device, model.dtype
@@ -399,7 +404,9 @@ def test_small_model_integration_test_llama_batched_regression(self):
             "USER: <|img|>\nWhat is this?\nASSISTANT: Two cats lying on a bed!\nUSER: <|img|>\nAnd this?\nASSISTANT:",
         ]
         image1 = Image.open(requests.get(IMAGE_OF_VIEW_URL, stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=[image1, image2, image1], text=prompts, return_tensors="pt", padding=True)
         inputs = inputs.to(model.device, model.dtype)
diff --git a/tests/models/deepseek_vl_hybrid/test_image_processing_deepseek_vl_hybrid.py b/tests/models/deepseek_vl_hybrid/test_image_processing_deepseek_vl_hybrid.py
index f8646f1965cc..b2e4f6452031 100644
--- a/tests/models/deepseek_vl_hybrid/test_image_processing_deepseek_vl_hybrid.py
+++ b/tests/models/deepseek_vl_hybrid/test_image_processing_deepseek_vl_hybrid.py
@@ -20,6 +20,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -203,14 +204,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        import io
-
-        import httpx
-
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         encodings = {}
diff --git a/tests/models/eomt/test_modeling_eomt.py b/tests/models/eomt/test_modeling_eomt.py
index 54126d707f76..6dd03fddad10 100644
--- a/tests/models/eomt/test_modeling_eomt.py
+++ b/tests/models/eomt/test_modeling_eomt.py
@@ -15,25 +15,20 @@
 
 import unittest
 
-import requests
-
 from transformers import AutoImageProcessor, EomtConfig, EomtForUniversalSegmentation, pipeline
 from transformers.testing_utils import require_torch, require_torch_accelerator, require_torch_fp16, slow, torch_device
-from transformers.utils import is_torch_available, is_vision_available
+from transformers.utils import is_torch_available
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor
 from ...test_pipeline_mixin import PipelineTesterMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
     import torch
 
 
-if is_vision_available():
-    from PIL import Image
-
-
 class EomtForUniversalSegmentationTester:
     def __init__(
         self,
@@ -175,7 +170,9 @@ def test_inference(self):
         model = EomtForUniversalSegmentation.from_pretrained(self.model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -218,7 +215,9 @@ def test_inference_fp16(self):
         model = EomtForUniversalSegmentation.from_pretrained(self.model_id, dtype=torch.float16, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -234,7 +233,9 @@ def test_semantic_segmentation_inference(self):
         model = EomtForUniversalSegmentation.from_pretrained(model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -271,7 +272,9 @@ def test_panoptic_segmentation_inference(self):
         model = EomtForUniversalSegmentation.from_pretrained(self.model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -320,7 +323,9 @@ def test_instance_segmentation_inference(self):
         model = EomtForUniversalSegmentation.from_pretrained(model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -366,7 +371,9 @@ def test_instance_segmentation_inference(self):
 
     @slow
     def test_segmentation_pipeline(self):
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         pipe = pipeline(model=self.model_id, subtask="panoptic", device=torch_device)
         output = pipe(image)
diff --git a/tests/models/eomt_dinov3/test_modeling_eomt_dinov3.py b/tests/models/eomt_dinov3/test_modeling_eomt_dinov3.py
index 1e3e2f872dde..7c708015063a 100644
--- a/tests/models/eomt_dinov3/test_modeling_eomt_dinov3.py
+++ b/tests/models/eomt_dinov3/test_modeling_eomt_dinov3.py
@@ -15,8 +15,6 @@
 
 import unittest
 
-import requests
-
 from transformers import AutoImageProcessor, EomtDinov3Config, EomtDinov3ForUniversalSegmentation, pipeline
 from transformers.testing_utils import (
     Expectations,
@@ -26,21 +24,18 @@
     slow,
     torch_device,
 )
-from transformers.utils import is_torch_available, is_vision_available
+from transformers.utils import is_torch_available
 
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, _config_zero_init, floats_tensor
 from ...test_pipeline_mixin import PipelineTesterMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
     import torch
 
 
-if is_vision_available():
-    from PIL import Image
-
-
 class EomtDinov3ForUniversalSegmentationTester:
     def __init__(
         self,
@@ -220,7 +215,9 @@ def test_inference(self):
         model = EomtDinov3ForUniversalSegmentation.from_pretrained(self.model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -262,7 +259,9 @@ def test_inference_bf16(self):
         )
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -319,7 +318,9 @@ def test_semantic_segmentation_inference(self):
         model = EomtDinov3ForUniversalSegmentation.from_pretrained(model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -375,7 +376,9 @@ def test_panoptic_segmentation_inference(self):
         model = EomtDinov3ForUniversalSegmentation.from_pretrained(self.model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -421,7 +424,9 @@ def test_instance_segmentation_inference(self):
         model = EomtDinov3ForUniversalSegmentation.from_pretrained(model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(model_id)
 
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=image, return_tensors="pt").to(model.device)
 
@@ -463,7 +468,9 @@ def test_instance_segmentation_inference(self):
             self.assertTrue(0.0 <= info["score"] <= 1.0)
 
     def test_segmentation_pipeline(self):
-        image = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         model = EomtDinov3ForUniversalSegmentation.from_pretrained(self.model_id, device_map="auto")
         processor = AutoImageProcessor.from_pretrained(self.model_id)
diff --git a/tests/models/fast_vlm/test_modeling_fast_vlm.py b/tests/models/fast_vlm/test_modeling_fast_vlm.py
index 5e26b591f339..d9b327d7eb75 100644
--- a/tests/models/fast_vlm/test_modeling_fast_vlm.py
+++ b/tests/models/fast_vlm/test_modeling_fast_vlm.py
@@ -39,6 +39,7 @@
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -282,7 +283,9 @@ def test_small_model_integration_test_batch(self):
             "user\n<image>\nWhat is this?\nassistant",
         ]
         image1 = Image.open(requests.get("https://llava-vl.github.io/static/images/view.jpg", stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         self.processor.tokenizer.padding_side = "left"
         inputs = self.processor(images=[image1, image2], text=prompts, return_tensors="pt", padding=True).to(
diff --git a/tests/models/flava/test_image_processing_flava.py b/tests/models/flava/test_image_processing_flava.py
index cb0dd10b6b50..dafab495a483 100644
--- a/tests/models/flava/test_image_processing_flava.py
+++ b/tests/models/flava/test_image_processing_flava.py
@@ -12,18 +12,16 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import io
 import random
 import unittest
 
-import httpx
 import numpy as np
-from PIL import Image
 
 from transformers.testing_utils import require_torch, require_vision
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -404,10 +402,8 @@ def test_slow_fast_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/models/fuyu/test_image_processing_fuyu.py b/tests/models/fuyu/test_image_processing_fuyu.py
index f0b1a3110c01..2aa89b54a493 100644
--- a/tests/models/fuyu/test_image_processing_fuyu.py
+++ b/tests/models/fuyu/test_image_processing_fuyu.py
@@ -1,7 +1,5 @@
-import io
 import unittest
 
-import httpx
 import numpy as np
 import pytest
 
@@ -17,6 +15,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available() and is_vision_available():
@@ -180,10 +179,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         encodings = {}
diff --git a/tests/models/idefics/test_image_processing_idefics.py b/tests/models/idefics/test_image_processing_idefics.py
index c3b646543a2d..9acba9a18c18 100644
--- a/tests/models/idefics/test_image_processing_idefics.py
+++ b/tests/models/idefics/test_image_processing_idefics.py
@@ -11,10 +11,8 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import io
 import unittest
 
-import httpx
 import pytest
 
 from transformers.testing_utils import (
@@ -28,6 +26,7 @@
 from transformers.utils import is_torch_available, is_torchvision_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -151,10 +150,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/models/imagegpt/test_image_processing_imagegpt.py b/tests/models/imagegpt/test_image_processing_imagegpt.py
index 8853b12381d7..4ac6b3bc150f 100644
--- a/tests/models/imagegpt/test_image_processing_imagegpt.py
+++ b/tests/models/imagegpt/test_image_processing_imagegpt.py
@@ -20,7 +20,6 @@
 
 import numpy as np
 import pytest
-import requests
 from datasets import load_dataset
 
 from transformers import AutoImageProcessor
@@ -35,6 +34,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -285,8 +285,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         encodings = {}
diff --git a/tests/models/kimi_k25/test_image_processing_kimi_k25.py b/tests/models/kimi_k25/test_image_processing_kimi_k25.py
index 8e973825c33a..927fa45d5645 100644
--- a/tests/models/kimi_k25/test_image_processing_kimi_k25.py
+++ b/tests/models/kimi_k25/test_image_processing_kimi_k25.py
@@ -12,10 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import io
 import unittest
 
-import httpx
 import numpy as np
 
 from transformers.image_utils import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
@@ -23,6 +21,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs, prepare_video_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -267,10 +266,8 @@ def test_custom_pixels(self):
 
     # Override to test additional outputs for equivalence such as `image_grid_thw`
     def test_backends_equivalence(self):
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/models/layoutlmv2/test_image_processing_layoutlmv2.py b/tests/models/layoutlmv2/test_image_processing_layoutlmv2.py
index 13f24bfda38b..d91c82769308 100644
--- a/tests/models/layoutlmv2/test_image_processing_layoutlmv2.py
+++ b/tests/models/layoutlmv2/test_image_processing_layoutlmv2.py
@@ -15,7 +15,6 @@
 import unittest
 
 import pytest
-import requests
 
 from transformers.testing_utils import (
     require_pytesseract,
@@ -26,19 +25,16 @@
     torch_device,
 )
 from transformers.utils import (
-    is_pytesseract_available,
     is_torch_available,
 )
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
     import torch
 
-if is_pytesseract_available():
-    from PIL import Image
-
 
 class LayoutLMv2ImageProcessingTester:
     def __init__(
@@ -148,8 +144,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         encodings = {}
diff --git a/tests/models/lfm2_vl/test_modeling_lfm2_vl.py b/tests/models/lfm2_vl/test_modeling_lfm2_vl.py
index 74e34cb9ea4d..746433978bc0 100644
--- a/tests/models/lfm2_vl/test_modeling_lfm2_vl.py
+++ b/tests/models/lfm2_vl/test_modeling_lfm2_vl.py
@@ -37,6 +37,7 @@
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
+from ...test_processing_common import load_test_image
 
 
 if is_vision_available():
@@ -210,8 +211,8 @@ class Lfm2VlForConditionalGenerationIntegrationTest(unittest.TestCase):
     def setUp(self):
         self.processor = AutoProcessor.from_pretrained("LiquidAI/LFM2-VL-1.6B")
         self.processor.tokenizer.padding_side = "left"
-        self.image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        self.image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
         self.image2 = Image.open(
             BytesIO(
@@ -296,8 +297,8 @@ class Lfm2_5VlForConditionalGenerationIntegrationTest(unittest.TestCase):
     def setUp(self):
         self.processor = AutoProcessor.from_pretrained("LiquidAI/LFM2.5-VL-1.6B")
         self.processor.tokenizer.padding_side = "left"
-        self.image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        self.image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
         self.image2 = Image.open(
             BytesIO(
diff --git a/tests/models/llava/test_modeling_llava.py b/tests/models/llava/test_modeling_llava.py
index b0bcf5afbbbd..3fb9446ef68d 100644
--- a/tests/models/llava/test_modeling_llava.py
+++ b/tests/models/llava/test_modeling_llava.py
@@ -45,6 +45,7 @@
 from ...test_configuration_common import ConfigTester
 from ...test_modeling_common import ModelTesterMixin, floats_tensor, ids_tensor
 from ...test_pipeline_mixin import PipelineTesterMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -367,7 +368,9 @@ def test_small_model_integration_test_llama_batched(self):
             "USER: <image>\nWhat is this? ASSISTANT:",
         ]
         image1 = Image.open(requests.get("https://llava-vl.github.io/static/images/view.jpg", stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=[image1, image2], text=prompts, return_tensors="pt", padding=True).to(torch_device)
 
@@ -418,7 +421,9 @@ def test_small_model_integration_test_batch(self):
             "USER: <image>\nWhat is this?\nASSISTANT:",
         ]
         image1 = Image.open(requests.get("https://llava-vl.github.io/static/images/view.jpg", stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = self.processor(images=[image1, image2], text=prompts, return_tensors="pt", padding=True).to(
             torch_device, torch.float16
@@ -471,7 +476,9 @@ def test_small_model_integration_test_llama_batched_regression(self):
             "USER: <image>\nWhat is this?\nASSISTANT: Two cats lying on a bed!\nUSER: <image>\nAnd this?\nASSISTANT:",
         ]
         image1 = Image.open(requests.get("https://llava-vl.github.io/static/images/view.jpg", stream=True).raw)
-        image2 = Image.open(requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw)
+        image2 = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
+        )
 
         inputs = processor(images=[image1, image2, image1], text=prompts, return_tensors="pt", padding=True).to(
             torch_device
diff --git a/tests/models/minimax_m3_vl/test_modeling_minimax_m3_vl.py b/tests/models/minimax_m3_vl/test_modeling_minimax_m3_vl.py
index ce483eeeffad..2047c20001cd 100644
--- a/tests/models/minimax_m3_vl/test_modeling_minimax_m3_vl.py
+++ b/tests/models/minimax_m3_vl/test_modeling_minimax_m3_vl.py
@@ -45,6 +45,7 @@
     ids_tensor,
 )
 from ...test_pipeline_mixin import PipelineTesterMixin
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -696,8 +697,8 @@ def test_padding_sides_text_and_image(self):
         # canonical COCO two-cats photo), paired with deliberately different-length questions so the
         # batch genuinely needs padding.
         dog = Image.open(requests.get("https://picsum.photos/id/237/400/300", stream=True).raw).convert("RGB")
-        cats = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        cats = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         ).convert("RGB")
         texts = [
             self._prompt(processor, "What animal is this? One word."),
diff --git a/tests/models/pix2struct/test_image_processing_pix2struct.py b/tests/models/pix2struct/test_image_processing_pix2struct.py
index 386de5e8527e..9a7a8a60e42d 100644
--- a/tests/models/pix2struct/test_image_processing_pix2struct.py
+++ b/tests/models/pix2struct/test_image_processing_pix2struct.py
@@ -22,7 +22,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
-from ...test_processing_common import url_to_local_path
+from ...test_processing_common import load_test_image, url_to_local_path
 
 
 if is_torch_available():
@@ -99,15 +99,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        import io
-
-        import httpx
-        from PIL import Image
-
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/models/qwen2_vl/test_image_processing_qwen2_vl.py b/tests/models/qwen2_vl/test_image_processing_qwen2_vl.py
index fdad0a9a19fc..907f4bb4366a 100644
--- a/tests/models/qwen2_vl/test_image_processing_qwen2_vl.py
+++ b/tests/models/qwen2_vl/test_image_processing_qwen2_vl.py
@@ -12,13 +12,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import io
 import itertools
 import json
 import tempfile
 import unittest
 
-import httpx
 import numpy as np
 
 from transformers.image_utils import OPENAI_CLIP_MEAN, OPENAI_CLIP_STD
@@ -27,6 +25,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs, prepare_video_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -294,10 +293,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/models/video_llama_3/test_image_processing_video_llama_3.py b/tests/models/video_llama_3/test_image_processing_video_llama_3.py
index b6d956e258b0..18c5d6d968d5 100644
--- a/tests/models/video_llama_3/test_image_processing_video_llama_3.py
+++ b/tests/models/video_llama_3/test_image_processing_video_llama_3.py
@@ -18,7 +18,6 @@
 import unittest
 
 import numpy as np
-import requests
 
 from transformers.image_utils import IMAGENET_STANDARD_MEAN, IMAGENET_STANDARD_STD
 from transformers.models.video_llama_3.image_processing_video_llama_3 import smart_resize
@@ -26,6 +25,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs, prepare_video_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -296,8 +296,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            requests.get("http://images.cocodataset.org/val2017/000000039769.jpg", stream=True).raw
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         encodings = {}
diff --git a/tests/models/vitpose/test_image_processing_vitpose.py b/tests/models/vitpose/test_image_processing_vitpose.py
index a20426d77249..5ce1a51b2eff 100644
--- a/tests/models/vitpose/test_image_processing_vitpose.py
+++ b/tests/models/vitpose/test_image_processing_vitpose.py
@@ -11,10 +11,8 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-import io
 import unittest
 
-import httpx
 import numpy as np
 import pytest
 
@@ -22,6 +20,7 @@
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_image_processing_common import ImageProcessingTestMixin, prepare_image_inputs
+from ...test_processing_common import load_test_image
 
 
 if is_torch_available():
@@ -233,10 +232,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
         boxes = [[[0, 0, 1, 1]]]
 
diff --git a/tests/test_image_processing_common.py b/tests/test_image_processing_common.py
index 4233df645d1a..75a1ba09eaec 100644
--- a/tests/test_image_processing_common.py
+++ b/tests/test_image_processing_common.py
@@ -12,7 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import inspect
-import io
 import json
 import os
 import pathlib
@@ -21,7 +20,6 @@
 import warnings
 from copy import deepcopy
 
-import httpx
 import numpy as np
 import pytest
 
@@ -41,6 +39,8 @@
 )
 from transformers.utils import is_torch_available, is_vision_available
 
+from .test_processing_common import load_test_image
+
 
 if is_torch_available():
     import torch
@@ -175,10 +175,8 @@ def test_backends_equivalence(self):
         if len(self.image_processing_classes) < 2:
             self.skipTest(reason="Skipping backends equivalence test as there are less than 2 backends")
 
-        dummy_image = Image.open(
-            io.BytesIO(
-                httpx.get("http://images.cocodataset.org/val2017/000000039769.jpg", follow_redirects=True).content
-            )
+        dummy_image = load_test_image(
+            "https://huggingface.co/datasets/hf-internal-testing/fixtures-coco/resolve/main/val2017/000000039769.jpg"
         )
 
         # Create processors for each backend
diff --git a/tests/test_processing_common.py b/tests/test_processing_common.py
index 8e756923c5e4..4fdb455d253f 100644
--- a/tests/test_processing_common.py
+++ b/tests/test_processing_common.py
@@ -26,6 +26,7 @@
 from huggingface_hub import hf_hub_download
 from parameterized import parameterized
 
+from transformers.image_utils import load_image
 from transformers.processing_utils import (
     MODALITY_TO_AUTOPROCESSOR_MAPPING,
     Unpack,
@@ -72,6 +73,10 @@
     MODALITY_INPUT_DATA[modality] = [url_to_local_path(url) for url in urls]
 
 
+def load_test_image(url):
+    return load_image(url_to_local_path(url))
+
+
 def prepare_image_inputs():
     """This function prepares a list of PIL images"""
     image_inputs = [np.random.randint(255, size=(3, 30, 400), dtype=np.uint8)]
diff --git a/utils/fetch_hub_objects_for_ci.py b/utils/fetch_hub_objects_for_ci.py
index dfab2e2cd786..bf6fd4d8b5d9 100644
--- a/utils/fetch_hub_objects_for_ci.py
+++ b/utils/fetch_hub_objects_for_ci.py
@@ -51,19 +51,6 @@
     "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_001.png",
     "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_001.png",
     "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/table_recognition.jpg",
-    # Don't use the original COCO URLs anymore. Replace with images from https://huggingface.co/datasets/hf-internal-testing/fixtures-coco below
-    "http://images.cocodataset.org/val2017/000000000139.jpg",
-    "http://images.cocodataset.org/val2017/000000000285.jpg",
-    "http://images.cocodataset.org/val2017/000000000632.jpg",
-    "http://images.cocodataset.org/val2017/000000000724.jpg",
-    "http://images.cocodataset.org/val2017/000000000776.jpg",
-    "http://images.cocodataset.org/val2017/000000000785.jpg",
-    "http://images.cocodataset.org/val2017/000000000802.jpg",
-    "http://images.cocodataset.org/val2017/000000000872.jpg",
-    "http://images.cocodataset.org/val2017/000000001000.jpg",
-    "http://images.cocodataset.org/val2017/000000039769.jpg",
-    "http://images.cocodataset.org/val2017/000000077595.jpg",
-    "http://images.cocodataset.org/val2017/000000136466.jpg",
     "https://cdn.britannica.com/59/94459-050-DBA42467/Skyline-Chicago.jpg",
     "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg",
     "https://llava-vl.github.io/static/images/view.jpg",