SkillAgentSearch skills...

ResNet FPN Object Detection

Built a Feature Pyramid Network with a ResNet-50 backbone with a custom Torch loss function for rotated bounding boxes. Used it for object detection and segmentation on adversarial noisy binary images of stars, achieving over 96% IOU accuracy.

Install / Use

npx skills add yashpansari/ResNet-FPN-Object-Detection

Installs into whichever agent you are using.

About this skill

Quality Score

0/100

Supported Platforms

Universal

README

ResNet-FPN-Object-Detection

Built a Feature Pyramid Network with a ResNet-50 backbone with a custom Torch loss function for rotated bounding boxes. Used it for object detection and segmentation on adversarial noisy binary images of stars, achieving over 96% IOU accuracy wit under 1.5 Million parameters. Trained on 1 standard Tesla GPU for 5 hours (150 epochs).

Papers cited within code.

The data synthesizer generates images and labels. The model determines whether each image has a star and, if so, finds a rotated bounding box that bounds the star.

More precisely, the labels contain the following five numbers, which the model predicts:

  • the x and y coordinates of the center
  • yaw
  • width and height.

The model is evaluated quantitatively using compute_score.py. The metric is the percent of correctly identified stars based on an IOU (Intersection of Union) threshold of 0.7 (for 1024 random samples).

Related Skills

View on GitHub
GitHub Stars6
CategoryDevelopment
Updated2y ago
Forks0

Languages

Python

Security Score

70/100

Audited on Jul 30, 2024

No findings