BASNET model created using tensorflow.
Project description
SOD-BASNET-Py
BASNet: Boundary-Aware Salient Object Detection
- Deep Convolutional Neural Networks have been adopted for salient object detection and achieved the state-of-the-art performance. Most of the previous works however focus on region accuracy but not on the boundary quality.
- In this project, we propose a predict-refine architecture, BASNet, and a new hybrid loss for Boundary-Aware Salient object detection. Specifically, the architecture is composed of a densely supervised Encoder-Decoder network and a residual refinement module, which are respectively in charge of saliency prediction and saliency map refinement.
- The Research paper referred is BASNet: Boundary-Aware Salient Object Detection
Objective
- The Official Code for the Basnet Model is provided by the authors on the github.
- The Official Deep Learning model is made using Pytorch , my Goal is to create a similar model using tensorflow.
- Why ? Whilst Pytorch provide better developer experience and error handeling, I find Tensorflow to be a great ML Framework to work with for Beginners .
- So the objective for this repo is to create Basnet Model using Tensorflow.
Architecture
- The Architecture Proposed by the authors is predict-refine architecture.
- The Author have used Transfer Learning to improve model performance . They Used First 4 layers of the resnet34 model which is pretrained in imagenet dataset.
- But for the sake of understanding abstract architecture of the Basnet model I haven't used the transfer learning method , instead created each layer individually.
Loss
- The hybrid loss guides the network to learn the transformation between the input image and the ground truth in a three-level hierarchy – pixel-, patch- and map- level – by fusing Binary Cross Entropy (BCE), Structural SIMilarity (SSIM) and Intersectionover- Union (IoU) losses.
Thank You
- This is my attempt to build the Basnet model using the research paper . If you find any error or mistakes let me know . Love all the feedbacks.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
SODBASNET-0.0.5.tar.gz
(6.1 kB
view hashes)
Built Distribution
Close
Hashes for SODBASNET-0.0.5-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6d9e330766785f2c77895ab017dc4407f2fecf931db6bcec4cc4e9d4483db6e5 |
|
MD5 | d6219bf65bb75e7cbabf90185d4351e1 |
|
BLAKE2b-256 | bbd971d95615298747327165c79f95c42bd93c6ea35bfac7d91334f7b04514ed |