Deep Neural Networks Achieve Best Classification Accuracy On Videos. However, Traditional Methods Or Shallow Architectures Remain Competitive And Combinations Of Different Network Types Are The Usual Chosen Approach. A Reason For This Less Important Impact Of Deep Methods For Video Recognition Is The Motion Representation. The Time Has A Stronger Redundancy, And An Important Elasticity Compared To The Spatial Dimensions. The Temporal Redundancy Is Evident, But The Elasticity Within An Action Class Is Well Less Considered .Several Instances Of The Action Still Widely Differ By Their Style And Speed Of Execution.
Road Accidents Are Man-made Cataclysmic Phenomena And Are Not Generally Predictable. With Increasing Numbers Of Deaths Due To Accidents In The Roadways, A Smart And Fast Detection System For Road Accidents Is The Need Of The Hour. Often, Precious Few Seconds After The Accidents Make The Difference Between Life And Death. To Address This Problem More Efficiently, “A Novel Approach For Road Accident Detection In CCTV Videos Using DETR Algorithm” Has Been Developed To Aid In Notifying Hospitals And The Local Police At Places Where Instant Notification Is Seldom Feasible. This Paper Presents A Novel And Efficient Method For Detecting Road Accidents With DETR (Detection Transformers) And Random Forest Classifier. Objects Such As Cars, Bikes, People, Etc. In The CCTV Footage Are Detected Using The DETR And The Features Are Fed To A Random Forest Classifier For Frame Wise Classification. Each Frame Of The Video Is Classified As An Accident Frame Or A Non-accident Frame. A Total Count Of Predicted Accident Frames From Any 60 Continuous Frames Of The Video Are Considered Using A Sliding Window Technique Before The Final Decision Is Made. Simulation Results Show That The Proposed System Achieves 78.2% Detection Rate In CCTV Videos.
Human Pose Estimation In Images Is Challenging And Important For Many Computer Vision Applications. Large Improvements In Human Pose Estimation Have Been Achieved With The Development Of Convolutional Neural Networks. Even Though, When Encountered Some Difficult Cases Even The State-of-the-art Models May Fail To Predict All The Body Joints Correctly. Some Recent Works Try To Refine The Pose Estimator. GAN (Generative Adversarial Networks) Has Been Proved To Be Efficient To Improve Human Pose Estimation. However, GAN Can Only Learn Local Body Joints Structural Constrains. In This Paper, We Propose To Apply Self-Attention GAN To Further Improve The Performance Of Human Pose Estimation. With Attention Mechanism In The Framework Of GAN, We Can Learn Long-range Body Joints Dependencies, Therefore Enforce The Entire Body Joints Structural Constrains To Make All The Body Joints To Be Consistent. Our Method Outperforms Other State-of-the-art Methods On Two Standard Benchmark Datasets MPII And LSP For Human Pose Estimation.