project_coonhound/anima_module.yaml at main · RobotFlow-Labs/project_coonhound · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
name: coonhound
codename: COONHOUND
wave: 10
theme: WARDOG
version: 0.3.0

paper:
  title: "MambaNUT: Nighttime UAV Tracking via Mamba-based Adaptive Curriculum Learning"
  arxiv: "2412.00626"
  venue: "IROS 2025"
  authors: "You Wu, Xiangyang Yang, Xucheng Wang, Hengzhou Ye, Dan Zeng, Shuiwang Li"

model:
  architecture: "Vision Mamba + Center Head"
  backbone: mambar_small_patch16_224
  embed_dim: 384
  depth: 24
  params_M: 4.1
  flops_GMAC: 1.1
  template_size: 128
  search_size: 256
  head_type: CENTER

training:
  optimizer: AdamW
  lr: 0.0004
  weight_decay: 0.0001
  epochs: 300
  batch_size: 32
  scheduler: cosine_warmup
  losses: [giou, l1, adw]

deployment:
  backends: [cuda, mlx, cpu]
  serve_port: 8080
  endpoints: [/health, /ready, /predict, /track, /reset, /info]
  export_formats: [safetensors, onnx, tensorrt_fp16, tensorrt_fp32]

dependencies:
  detector: yolo26m-uav
  pretrained: mambar_small_patch16_224.pth
  datasets:
    - name: LaSOT
      role: daytime_train
    - name: GOT-10k
      role: daytime_train
    - name: COCO2017
      role: daytime_train
    - name: TrackingNet
      role: daytime_train
    - name: BDD100K-Night
      role: nighttime_train
    - name: SHIFT-Night
      role: nighttime_train
    - name: ExDark
      role: nighttime_train
    - name: NAT2024-1
      role: evaluation
    - name: NAT2021
      role: evaluation
    - name: UAVDark135
      role: evaluation

benchmarks:
  - name: NAT2024-1
    metric: precision
    paper: 83.3
    gate: 79.0
  - name: NAT2021
    metric: precision
    paper: 70.1
    gate: 66.0
  - name: UAVDark135
    metric: precision
    paper: 70.0
    gate: 66.0

status:
  code_complete: true
  training_complete: false
  evaluation_complete: false
  export_complete: false
  mvp_readiness: 60