Pythonic AI
493 subscribers
317 photos
84 videos
67 files
538 links
منابع، دوره ها، همایشها ، مقالات و میم کامپیوتر
Download Telegram
Orientation Aware Object Detection with Application to Firearms

To address the problems in axis-aligned bounding boxes like clutter and redundant features in the background, we propose orientation aware design. Our proposed method (OAOD) separates the foreground information from background information effectively, yields more accuracy than axis-aligned bounding boxes based algorithms (Faster RCNN, SSD, Yolo, DSSD). Without initializing oriented proposals, we present multi-stage detector training using angle information only along with location points and class labels as ground truth. OAOD is evidently more effective in case of a long object like Rifle in firearms.

DATASET: Link
Trained Model: Link
The paper is available on arXiv: Link

http://im.itu.edu.pk/orientation-aware-firearms-detection/

https://github.com/makhtar17004/orientation-aware-firearm-detection

#machinelearning #deeplearning #artificialintelligence #detection

@pythonicAI