How altzits007/torchvprsnn works
This repository implements visual place recognition with a rate-encoded spiking neural network (SNN). The principal workflow loads reference and query images, encodes them as spike trains, trains or loads an SNN, assigns neurons to places, and evaluates query matches, including sequential aggregation. The graph includes documented experiment and reporting branches. Some runtime details are unsampled; edges are limited to relationships supported by excerpts, the source index, or README statements.
Main entry points: snn_model.py Training & Inference · networks.py SNN Network · encoders.py Rate Encoding · run_table_iv_real.py Nordland Ablation
Components 17
VPR Workflow
- Training & Inference
core/snn_model.py - Dataset Loading
core/vprsnn_evaluation.py - Prediction API
core/predict.py - Model Checkpoints
core/snn_model.py - Runtime Profiling
core/vprsnn_evaluation.py
SNN Runtime
- Rate Encoding
core/encoders.py - SNN Network
core/networks.py - Learning Mechanisms
core/mechanisms.py
Assignments & Evaluation
- Neuron Assignments
core/neuronal_assignments.py - Place Recognition
core/vprsnn_evaluation.py - VPR Metrics
core/metrics.py
Experiments & Reporting
- Nordland Ablation
experiments/run_table_iv_real.py - Oxford Ablation
experiments/run_table_iv_oxford.py - Window Study
experiments/run_table_v.py - R@100P Reports
tools/plot_r100p_from_extracted_tables.py
Other components
- Experimenter
- Image Datasets
Connections 31
- Experimenter → Training & Inference: runs
README.md - Image Datasets → Dataset Loading: provides images
README.md - Training & Inference → Dataset Loading: loads data
core/snn_model.py - Training & Inference → Rate Encoding: configures
core/snn_model.py - Training & Inference → SNN Network: initializes
core/snn_model.py - Rate Encoding → SNN Network: encodes input
README.md - SNN Network → Learning Mechanisms: uses learning rules
core/networks.py - Training & Inference → Neuron Assignments: assigns neurons
core/snn_model.py - Training & Inference → Place Recognition: evaluates places
core/snn_model.py - Place Recognition → VPR Metrics: computes PR metrics
core/vprsnn_evaluation.py - Training & Inference → Model Checkpoints: saves or loads
README.md - Training & Inference → Runtime Profiling: profiles runtime
README.md - Experimenter → Prediction API: requests prediction
- Prediction API → Rate Encoding: uses
core/predict.py - Prediction API → SNN Network: uses
core/predict.py - Prediction API → Place Recognition: uses
core/predict.py - Experimenter → Nordland Ablation: runs study
README.md - Experimenter → Oxford Ablation: runs study
README.md - Experimenter → Window Study: runs study
README.md - Nordland Ablation → Rate Encoding: encodes images
experiments/run_table_iv_real.py - Nordland Ablation → SNN Network: trains network
experiments/run_table_iv_real.py - Nordland Ablation → Neuron Assignments: builds assignments
experiments/run_table_iv_real.py - Nordland Ablation → Place Recognition: evaluates queries
experiments/run_table_iv_real.py - Oxford Ablation → Rate Encoding: encodes images
experiments/run_table_iv_oxford.py - Oxford Ablation → SNN Network: trains network
experiments/run_table_iv_oxford.py - Oxford Ablation → Neuron Assignments: builds assignments
experiments/run_table_iv_oxford.py - Oxford Ablation → Place Recognition: evaluates queries
experiments/run_table_iv_oxford.py - Window Study → Rate Encoding: encodes images
experiments/run_table_v.py - Window Study → SNN Network: trains network
experiments/run_table_v.py - Window Study → Neuron Assignments: aggregates scores
experiments/run_table_v.py - R@100P Reports → Experimenter: produces boxplots
README.md