Akhil Raj – Medium

Dec 14, 2022

Understanding FCOS: Fully Convolutional One-Stage Object Detection

Introduction WHAT A fully convolutional one-stage object detector (FCOS) to solve object detection in a per-pixel prediction fashion, analogue to semantic segmentation. WHY Almost all state-of-the-art object detectors such as RetinaNet, SSD, YOLOv3, and Faster R-CNN rely on pre-defined anchor boxes. In contrast, the proposed detector FCOS is anchor box…

Computer Vision

9 min read

Understanding FCOS: Fully Convolutional One-Stage Object Detection

Computer Vision

9 min read

Jun 25, 2021

Social LSTM: A Deep Learning Model that Predicts the Future of your Path Trajectory!

Note: This blog is an attempt to explain the research paper describing the Deep Learning model Social-LSTM. The paper can be found here. Why Social LSTM Social LSTM is a model that predicts the future trajectories of pedestrians (which we can customize for other kinds of objects’ trajectories) given the data of their…

Deep Learning

4 min read

Deep Learning

4 min read

Published in

MLearning.ai

·Mar 17, 2021

Understanding RetinaMask

This article is an attempt to explain RetinaMask Deep Learning Model as published in this research paper, but more simply and shortly. GitHub Repository is also available[7]. Introduction WHY To match the accuracy of One-stage detectors[7], which are extremely important and popular in applications where speed and computational resources are important design…

Deep Learning

7 min read

Deep Learning

7 min read

Mar 17, 2021

Understanding RetinaNet

Introduction WHY The extreme foreground-background class imbalance encountered during training of dense detectors is the central cause for one-stage detectors trailing the accuracy of two-stage detectors. HOW Authors propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples…

Deep Learning

5 min read

Deep Learning

5 min read

Published in

MLearning.ai

·Mar 17, 2021

Faster RCNN : Towards Real-Time Object Detection with Region Proposal Networks

Introduction WHY Region proposals are the test-time computational bottleneck in state-of-the-art detection systems. The region proposal step still consumes as much running time as the detection network. HOW We introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is…

Faster R Cnn

5 min read

Faster R Cnn

5 min read