This guide maps the utilities that currently exist under tools/.
| Tool | Main files | Purpose |
|---|---|---|
| quality comparator | tools/quality-comparator/comparator.py, visualizer.py |
compare SR outputs against HR references |
| video analyzer | tools/video-analyzer/analyzer.py, metrics.py, main.py |
analyze video quality and temporal behavior |
| benchmark runner | tools/benchmark-runner/runner.py, profiler.py |
run TFLite enhancement on desktop and collect timings |
| reporting | tools/reporting/report_generator.py, visualizer.py |
build reports and charts from benchmark CSVs |
| docx creator | tools/docx-creator/converter.py |
convert supported text sources into DOCX outputs |
See the full README:
../../tools/quality-comparator/README.md
Typical use:
python tools/quality-comparator/comparator.py --lr_dir inputs --sr_dir outputs --hr_dir references --output_dir outputs/comparison
python tools/quality-comparator/visualizer.py --lr input.png --sr span.png eduscale.png --hr reference.png --output outputs/comparison.pngSee the full README:
../../tools/video-analyzer/README.md
Typical use:
python tools/video-analyzer/analyzer.py --input video.mp4 --model models/span/original/span_x2_ch48.pth --output_dir outputs/video-analysisThis is the desktop TFLite runner used by the benchmark scripts.
python tools/benchmark-runner/runner.py --input datasets/benchmark/lr --output_dir outputs/enhanced --preset balancedThe runner can handle both image directories and individual videos. It also
writes inference-times.csv into the output directory unless you override it.
Generate summary docs from benchmarks/results/:
python tools/reporting/report_generator.py --results_dir benchmarks/results --output_format markdown --output benchmarks/results/report.mdUse tools/reporting/visualizer.py when you want plots rather than text
reports.
The tools/docx-creator/ folder includes a converter and its own README. Use
it when you need Word-formatted deliverables built from repo content or related
artifacts.
- generate or collect datasets
- train and export a checkpoint
- run desktop enhancement with the benchmark runner
- compare outputs with the quality comparator
- summarize results with reporting scripts
- explore trends in notebooks under
experiments/