Skip to content

Commit 39da1ba

Browse files
PatBall1cursoragent
andcommitted
Add conda-forge recipe for staged-recipes submission
Includes detectron2 + pytorch as run dependencies so that conda install detectree2 fully resolves the deep learning stack including CUDA support. Co-authored-by: Cursor <cursoragent@cursor.com>
1 parent f8885cf commit 39da1ba

1 file changed

Lines changed: 69 additions & 0 deletions

File tree

recipe/meta.yaml

Lines changed: 69 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,69 @@
1+
{% set version = "2.1.2" %}
2+
3+
package:
4+
name: detectree2
5+
version: {{ version }}
6+
7+
source:
8+
url: https://pypi.org/packages/source/d/detectree2/detectree2-{{ version }}.tar.gz
9+
sha256: 2c6a08a394682fd0674aaf9c1a7698dde56e834c8aba9629ae131063dcec8801
10+
11+
build:
12+
noarch: python
13+
script: {{ PYTHON }} -m pip install . -vv --no-deps --no-build-isolation
14+
number: 0
15+
16+
requirements:
17+
host:
18+
- python >=3.8
19+
- setuptools >=68.0
20+
- setuptools-scm >=8.0
21+
- pip
22+
run:
23+
- python >=3.8
24+
# Core
25+
- pyyaml >=5.1
26+
- numpy >=1.20
27+
- pandas >=1.3
28+
- tqdm >=4.60
29+
- opencv >=4.5
30+
# Geospatial
31+
- shapely >=2.0
32+
- geopandas >=0.13
33+
- rasterio >=1.2
34+
- fiona >=1.8
35+
- rtree >=0.9
36+
# Evaluation
37+
- pycocotools >=2.0.4
38+
# Deep learning (resolved automatically by conda, including CUDA)
39+
- pytorch >=1.8
40+
- torchvision >=0.9
41+
- detectron2
42+
43+
test:
44+
imports:
45+
- detectree2
46+
- detectree2.preprocessing.tiling
47+
- detectree2.models.outputs
48+
commands:
49+
- pip check
50+
- python -c "from detectree2 import __version__; assert __version__ == '{{ version }}'"
51+
requires:
52+
- pip
53+
54+
about:
55+
home: https://github.com/PatBall1/detectree2
56+
summary: Automatic tree crown delineation in aerial RGB imagery based on Mask R-CNN.
57+
description: |
58+
detectree2 is a Python package for automatic tree crown delineation
59+
in aerial RGB and multispectral imagery based on Mask R-CNN.
60+
Pre-trained models are available in the model_garden.
61+
doc_url: https://patball1.github.io/detectree2/
62+
dev_url: https://github.com/PatBall1/detectree2
63+
license: MIT
64+
license_family: MIT
65+
license_file: LICENSE
66+
67+
extra:
68+
recipe-maintainers:
69+
- PatBall1

0 commit comments

Comments
 (0)