Computer Vision Backend for Comic Book Cover Detection
Sergei Serafimovich
Context
Users upload photos of comic books taken in real-world conditions: angled shots, uneven lighting, multiple covers in a single image. These images are not suitable for direct processing and require computer vision preprocessing before they can be used elsewhere in the system.
What was built
I developed a computer vision backend focused on preparing comic book cover images for further use in the application.
The system handles:
Detection of one or multiple comic book covers in a single image
Accurate cropping of detected covers
Perspective correction and geometric normalization
Output of clean, standardized cover images
How It Works
Cover Detection
A computer vision model detects comic book covers in user-uploaded images, including cases with multiple or partially visible covers.
Geometric Processing
For each detected cover, the system refines boundaries and corrects perspective distortion.
Skewed or angled covers are transformed into flat, normalized images.
Final Output
The backend returns structured results: bounding boxes, corrected images, and confidence scores.
These outputs are designed to be consumed by downstream systems (e.g. visual search or recognition pipelines).
Results & Benefits
Reliable cover extraction from imperfect, real-world photos
Significantly cleaner inputs for downstream image-based workflows
Reduced need for manual user corrections
Why This Matters
This computer vision pipeline bridges the gap between raw user photos and usable visual data, enabling image-based features without requiring ideal input conditions.
Like this project
Posted Dec 15, 2025
I developed a computer vision backend focused on preparing comic book cover images for further use in the application.
Likes
0
Views
0
Timeline
Oct 13, 2025 - Oct 21, 2025
