project_borzoi/anima_module.yaml at main · RobotFlow-Labs/project_borzoi · 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
84
85
module:
  name: anima_borzoi
  codename: BORZOI
  wave: 10
  theme: WARDOG
  version: 0.3.0
  status: scaffold-complete

paper:
  title: "A Tri-Modal Dataset and a Baseline System for Tracking Unmanned Aerial Vehicles"
  arxiv: "2511.18344"
  benchmark: "https://xuefeng-zhu5.github.io/MM-UAV/"
  baseline_repo: "https://github.com/JJGU2291/MM-UAV-Benchmark"
  evaluation_repo: "https://github.com/JJGU2291/MM-UAV-Evaluation-ToolKit"

runtime:
  backends: ["cuda", "mlx", "cpu"]
  default_backend: "auto"
  python: ">=3.10"

entrypoints:
  cli: "python -m anima_borzoi"
  serve: "python -m anima_borzoi.serve"
  train_stage1: "python scripts/train_stage1.py"
  train_stage2: "python scripts/train_stage2.py"
  eval: "bash scripts/run_eval.sh"

http:
  host_env: "BORZOI_HOST"
  port_env: "BORZOI_PORT"
  default_port: 8080
  health: "/health"
  ready: "/ready"
  predict: "/predict"

inputs:
  - name: rgb
    modality: image
    dtype: uint8
    shape: [3, 640, 640]
  - name: ir
    modality: image
    dtype: uint8
    shape: [1, 640, 640]
  - name: event
    modality: image
    dtype: uint8
    optional: true

outputs:
  - name: rgb_detections
    schema: "[{xyxy, score, class_id}]"
  - name: ir_detections
    schema: "[{xyxy, score, class_id}]"
  - name: tracks
    schema: "[{track_id, last_box, last_score, hits, age}]"

docker:
  cuda: "docker/Dockerfile.cuda"
  mlx: "docker/Dockerfile.mlx"
  compose: "docker/docker-compose.yaml"

training:
  stage1:
    epochs: 100
    optimizer: sgd
    lr: 0.005
    batch_size: 32
  stage2:
    epochs: 50
    optimizer: adamw
    lr: 0.005
    batch_size: 32
    freeze_backbone: true

hardware_target:
  gpu: "NVIDIA L4 (23GB)"
  min_vram_gb: 14
  target_vram_gb: 16

status_flags:
  datasets_ready: false
  gpus_available: false
  paper_reproduced: false
  tests_passing: true