Dog/Cat Breed Image Classification/Localization/Segmentation with Transfer Learning

Published: February 08, 2023

Who: the solo project
When: Mar. 2020 - June 2020
Where: COGS225 (Image Recognition) solo project
What: Dog/Cat breeds images classification, localization, and segmentation with transfer learning, multitask learning, and cut-mix data augmentation.
- Dataset: Oxford Pet dataset (open to the public).
How: Python. TensorFlow 2.x OpenCV. Pre-trained weights SOTA models in TensorFlow. Cut-mix image augmentation. Learning rate scheduler.
Learn: All the class assignments were requested writing in PyTorch, but this project was written in TensorFlow 2.x. Transfer learning from TensorFlow. SOTA models, augmentation skills and learning rate control. Multitask learning. IoU. U-net structure for image segmentation.
The repository is now opened for the public.
- It will be back to the private in Aug. 2021. Please contact me if you want to see in detail.
Featured slides