Skip to content

README.md

Latest commit

 

History

History
53 lines (37 loc) · 1.67 KB

README.md

File metadata and controls

53 lines (37 loc) · 1.67 KB

SLEAP Training & Evaluation

Train and evaluate SLEAP pose estimation models using COCO-formatted annotations.

Features

  • Automatic COCO to SLEAP format conversion
  • Support for multiple training strategies
  • Cached preprocessing for faster iterations
  • NME evaluation with customizable normalization

Quick Start

# Train a model
python train_sleap_with_coco.py --profile baseline_large_rf.topdown.json --output_dir ./sleap_model

# Evaluate model performance
python evaluate_sleap_nme.py --model_dir ./sleap_model --test_json test.json

# List available training profiles
python train_sleap_with_coco.py --list-profiles

# Train with custom settings
python train_sleap_with_coco.py \
    --profile baseline_medium_rf.topdown.json \
    --epochs 50 \
    --input-scale 0.5

Training Profiles

SLEAP offers various pre-configured training strategies:

  • Receptive Fields: large_rf, medium_rf, small_rf - Control context size
  • Approaches:
    • topdown - Instance detection followed by pose estimation
    • bottomup - Keypoint detection followed by grouping
    • single - Single instance scenarios
    • centroid - Center-based instance detection

Example profiles: baseline_large_rf.topdown.json, baseline.centroid.json

Requirements

Install SLEAP following their installation guide.

Key Features

  • Automatic Conversion: Seamlessly converts COCO annotations to SLEAP format
  • Data Caching: Preprocessed data cached for efficient re-training
  • Flexible Architecture: Choose from multiple model architectures and training strategies
  • GPU Support: Optimized for CUDA-enabled TensorFlow