How riship21/gait-recognition-project works
This is a research baseline for closed-set gait identification: CASIA-B silhouette sequences are preprocessed into Gait Energy Images (GEIs), flattened into features, matched against an enrolled gallery, and evaluated by Rank-1 accuracy. Separate scripts audit sequence quality, compare matchers, and visualize same- and different-subject GEIs or distances. The graph focuses on these documented runtime workflows; the dataset itself is external, and no source-backed runtime database or service is present.
Main entry points: gei.py GEI generation · baseline_identification.py Baseline identification · compare_matchers.py Matcher comparison · audit_dataset.py Dataset quality audit
Components 15
Representation
- Silhouette preprocessing
src/preprocess.py - GEI generation
src/gei.py
Recognition
- Baseline identification
src/baseline_identification.py - Gallery features
src/baseline_identification.py - Nearest-neighbor matching
src/baseline_identification.py - Rank-1 evaluation
src/baseline_identification.py - Baseline predictions
results/tables/baseline_predictions.csv
Analysis
- Matcher comparison
src/compare_matchers.py - Matcher results
results/tables/matcher_comparison.csv - Dataset quality audit
src/audit_dataset.py - Frame-count table
results/tables/frame_counts_090.csv - GEI distance comparison
src/compare_distances.py - Same-subject GEIs
src/compare_same_subject.py - Different-subject GEIs
src/compare_different_subjects.py
Other components
- CASIA-B silhouettes
Connections 17
- CASIA-B silhouettes → GEI generation: reads frames
- GEI generation → Silhouette preprocessing: preprocesses frames
- Baseline identification → GEI generation: generates GEIs
- Baseline identification → Gallery features: builds gallery
- Gallery features → GEI generation: generates GEIs
- Baseline identification → Nearest-neighbor matching: matches probes
- Nearest-neighbor matching → Gallery features: compares features
- Baseline identification → Rank-1 evaluation: evaluates predictions
- Rank-1 evaluation → Baseline predictions: writes results
- Matcher comparison → Gallery features: builds gallery
- Matcher comparison → GEI generation: generates probe GEIs
- Matcher comparison → Matcher results: writes results
- Dataset quality audit → CASIA-B silhouettes: counts frames
- Dataset quality audit → Frame-count table: writes counts
- GEI distance comparison → GEI generation: generates GEIs
- Same-subject GEIs → GEI generation: generates GEIs
- Different-subject GEIs → GEI generation: generates GEIs