Ashish Patel 🇮🇳’s Post

𝗗𝗮𝘆-𝟯𝟮𝟬 𝗖𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀 𝗼𝗳 𝗔𝗹𝗶𝗯𝗮𝗯𝗮 𝗚𝗿𝗼𝘂𝗽 𝗵𝗮𝘀 𝗣𝘂𝗯𝗹𝗶𝘀𝗵𝗲𝗱 𝗽𝗮𝗽𝗲𝗿 𝗼𝗻 𝗢𝗮𝗱𝗧𝗥: 𝗢𝗻𝗹𝗶𝗻𝗲 𝗔𝗰𝘁𝗶𝗼𝗻 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗲𝗿𝘀 Follow me for a similar post: 🇮🇳 Ashish Patel ------------------------------------------------------------------- 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗙𝗮𝗰𝘁𝘀 : 🔸 Paper: 𝗢𝗮𝗱𝗧𝗥: 𝗢𝗻𝗹𝗶𝗻𝗲 𝗔𝗰𝘁𝗶𝗼𝗻 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗲𝗿𝘀 🔸 This paper is published at ICML 2021. 🔸 The purpose of online action detection is to correctly identify ongoing actions from streaming videos without any access to the future. Recently, this task has received increasing attention due to its great potential of diverse application prospects in real life, such as autonomous driving, video surveillance, anomaly detection, etc ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 Most recent approaches for online action detection tend to apply Recurrent Neural networks (RNN) to capture long-range temporal structure. However, RNN suffers from non-parallelism and gradient vanishing, hence it is hard to be optimized. In this paper, we propose a new encoder-decoder framework based on Transformers, named OadTR, to tackle these problems.  🔸 The encoder attached with a task token aims to capture the relationships and global interactions between historical observations. The decoder extracts auxiliary information by aggregating anticipated future clip representations. Therefore, OadTR can recognize current actions by encoding historical information and predicting future context simultaneously.  🔸 We extensively evaluate the proposed OadTR on three challenging datasets: HDD, TVSeries, and THUMOS14. The experimental results show that OadTR achieves higher training and inference speeds than current RNN based approaches, and significantly outperforms the state-of-the-art methods in terms of both mAP and mcAP. ------------------------------------------------------------------- #computervision #artificialintelligence #innovation -------------------------------------------------------------------