Have you heard of #Tube #CNN ?

Object or human detection in video is crucial for many applications.

It can also have useful applications such as in repetitive #manufacturing tasks to monitor and proactively prevent catastrophes.

Compared to images, video provides additional cues which can help to disambiguate the detection problem.

Here the authors attempt to learn discriminative models for the temporal evolution of object appearance and to use such models for object detection.

They introduced space-time tubes corresponding to temporal sequences of bounding boxes. They propose a TPN network where two CNN architectures for generating and classifying tubes, respectively, this helps maximize object recall.

The Tube-CNN then implements a tube-level object detector in the video. Our method improves state of the art on two large-scale datasets for object detection in video: HollywoodHeads and ImageNet VID. Tube models show particular advantages in difficult dynamic scenes.

Link to paper: https://lnkd.in/d3DW5Qe
Pytorch implementation of Tube-CNN : https://lnkd.in/dzxGdtt

#deeplearning #CNN #machinelearning #videoanalytics

✴️ @AI_Python_EN
🗣 @AI_Python_Arxiv
❇️ @AI_Python