Ashish Patel 🇮🇳’s Post

View profile for Ashish Patel 🇮🇳

Oracle105K followers

Day-12 Computer Vision Learning RetinaNet : Focal Loss for Dense Object Detection 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗳𝗮𝗰𝘁𝘀 🔸 RetineNet is a 2017 ICCV Best Student Paper Award paper with more than 500 citations. 🔸 The first author, Tsung-Yi Lin, has become Research Scientist at Google Brain when he was presenting RetinaNet in 2017 ICCV 🔸 Focal Loss First time introduced in this paper. ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://bit.ly/35xCHGk 𝗣𝘆𝘁𝗼𝗿𝗰𝗵 RetinaNet : https://bit.ly/2KaRQWO 𝘁𝗳-𝗸𝗲𝗿𝗮𝘀 RetinaNet : https://bit.ly/35vqm5q ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 RetinaNet is a one-stage object detection model that utilizes a focal loss function to address class imbalance during training. Focal loss applies a modulating term to the cross entropy loss in order to focus learning on hard negative examples. 🔸 RetinaNet is a single, unified network composed of a backbone network and two task-specific subnetworks. 🔸 The backbone is responsible for computing a convolutional feature map over an entire input image and is an off-the-self convolutional network. Note: More in comments #innovation #artificialintelligence #computervision

  • No alternative text description for this image

Very useful

Like
Reply
Sonu Kumar

Wipro2K followers

5y

Thanks for sharing

Divyesh Bhatt

Fiverr Pro2K followers

5y

This will help me

Focal Loss (FL) is an improved version of Cross-Entropy Loss (CE) that tries to handle the class imbalance problem by assigning more weights to hard or easily misclassified examples

  • No alternative text description for this image

Two-Stage Detectors Architecture of #Retinanet 🔸 In two-stage detectors such as Faster R-CNN, the first stage, region proposal network (RPN) narrows down the number of candidate object locations to a small number (e.g. 1–2k), filtering out most background samples. 🔸 At the second stage, classification is performed for each candidate object location. Sampling heuristics using fixed foreground-to-background ratio (1:3), or online hard example mining (OHEM) to select a small set of anchors (e.g., 256) for each minibatch. #technology #deeplearning #india

  • No alternative text description for this image
See more comments

To view or add a comment, sign in

Explore content categories