GPU-accelerated manga sheet grading powered by CUDA. Analyze density, edges, and contrast in milliseconds.
Three stages. One command. Instant results.
Point the CLI at a single manga page or an entire directory. Supports PNG, JPG, and TIFF formats.
Light density vectors, Sobel/Canny edge detection, and contrast entropy run in parallel. On CUDA hardware, these operations execute on the GPU.
A composite score from 0-100 maps to a letter grade (S through F). Optional heatmap overlays show exactly where issues are.
Every feature designed around manga print quality analysis.
CUDA kernels via CuPy for luminance histograms and Sobel gradients. Automatic CPU fallback when no GPU is present.
Grade entire scan directories in a single pass. JSON output for integration with automation pipelines.
Heatmap overlays show density distribution. Grade badges and vector arrows highlight problem areas at a glance.
Sub-second analysis on GPU. Composite scoring from density, edge, and contrast metrics weighted by print relevance.
Tune grade boundaries to match your print standards. Adjust metric weights for different manga styles.
Proprietary analysis mapping directional light changes across image regions. Catches subtle tonal imbalances.
$ neon-grade ./scans/ --overlay +------------------------------------------------------------+--------+--------+ | File | Score | Grade | +------------------------------------------------------------+--------+--------+ | ./scans/page001.png | 96.3 | S | | ./scans/page002.png | 87.1 | A | | ./scans/page003.png | 72.4 | B | +------------------------------------------------------------+--------+--------+ Overlay saved: ./scans/page001_graded.png Overlay saved: ./scans/page002_graded.png Overlay saved: ./scans/page003_graded.png $ neon-grade page01.png --json [ { "score": 96.3, "grade": "S", "breakdown": { "density": 94.8, "edge": 97.2, "contrast": 96.9 } } ]
Built on proven tools for image analysis and GPU computing.
Install from PyPI. Works out of the box on CPU. Add GPU support with one extra flag.
# Install (CPU)
$ pip install neon-speed-grader
# With GPU acceleration
$ pip install neon-speed-grader[gpu]
# Grade a manga page
$ neon-grade page01.png --overlay