what i am reading now?
\n\nAbstract— This paper proposes a computationally efficientapproach to detecting objects natively in 3D point clouds using convolutional neural networks (CNNs). In particular, this is achieved by leveraging a feature-centric voting scheme to implement novel convolutional layers which explicitly exploit the sparsity encountered in the input. To this end, we examine the trade-off between accuracy and speed for different architectures and additionally propose to use an L1 penalty on the filter activations to further encourage sparsity in the intermediate representations. To the best of our knowledge, this is the first work to propose sparse convolutional layers and L1 regularisation for efficient large-scale processing of 3D data.We demonstrate the efficacy of our approach on the KITTI object detection benchmark and show that Vote3Deep models with as few as three layers outperform the previous state of the art in both laser and laser-vision based approaches by margins of up to 40% while remaining highly competitive in terms of processing time.
\n\ni have downloaded the paper,interested readers can read full paper from here : https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/3D%20Deep%20Learning/vote3deep.pdf
Reference video link : https://www.youtube.com/watch?v=WUOSmAfeXIw
what i am reading now?
\n\nAbstract— This paper proposes a computationally efficientapproach to detecting objects natively in 3D point clouds using convolutional neural networks (CNNs). In particular, this is achieved by leveraging a feature-centric voting scheme to implement novel convolutional layers which explicitly exploit the sparsity encountered in the input. To this end, we examine the trade-off between accuracy and speed for different architectures and additionally propose to use an L1 penalty on the filter activations to further encourage sparsity in the intermediate representations. To the best of our knowledge, this is the first work to propose sparse convolutional layers and L1 regularisation for efficient large-scale processing of 3D data.We demonstrate the efficacy of our approach on the KITTI object detection benchmark and show that Vote3Deep models with as few as three layers outperform the previous state of the art in both laser and laser-vision based approaches by margins of up to 40% while remaining highly competitive in terms of processing time.
\n\ni have downloaded the paper,interested readers can read full paper from here : https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/3D%20Deep%20Learning/vote3deep.pdf
Reference video link : https://www.youtube.com/watch?v=WUOSmAfeXIw
Hi . Interesting . But it wil be applicable for Lidar type of data right ? Like SECOND ?
", "rawMarkdown": "Hi . Interesting . But it wil be applicable for Lidar type of data right ? Like SECOND ?", "votes": null }, { "id": "678514", "postDate": "11/21/2019 13:37:26", "content": "hello @phoenix9032 \ngood question,\nthe paper mentioned can be applicable with lidar type of data and that cnn can also be used for this kind of competition problem
", "rawMarkdown": "hello @phoenix9032 \ngood question,\nthe paper mentioned can be applicable with lidar type of data and that cnn can also be used for this kind of competition problem", "votes": null }, { "id": "680844", "postDate": "11/25/2019 10:18:25", "content": "Thank you for sharing! this is very interesting!
", "rawMarkdown": "Thank you for sharing! this is very interesting!", "votes": null }, { "id": "680845", "postDate": "11/25/2019 10:20:13", "content": "glad it helps
", "rawMarkdown": "glad it helps", "votes": null } ], "comments": [ { "id": 678354, "author_name": "phoenix9032", "author_url": "", "post_date": "11/21/2019 09:09:51", "content": "Hi . Interesting . But it wil be applicable for Lidar type of data right ? Like SECOND ?
", "votes": null, "replies": [ { "id": 678514, "author_name": "mobassir", "author_url": "", "post_date": "11/21/2019 13:37:26", "content": "hello @phoenix9032 \ngood question,\nthe paper mentioned can be applicable with lidar type of data and that cnn can also be used for this kind of competition problem
", "votes": null, "replies": [] } ] }, { "id": 680844, "author_name": "amelnozieres", "author_url": "", "post_date": "11/25/2019 10:18:25", "content": "Thank you for sharing! this is very interesting!
", "votes": null, "replies": [ { "id": 680845, "author_name": "mobassir", "author_url": "", "post_date": "11/25/2019 10:20:13", "content": "glad it helps
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "678335": "what i am reading now?\n# Paper Title : Vote3Deep: Fast Object Detection in 3D Point Clouds Using Efficient Convolutional Neural Networks\n\nAbstract— This paper proposes a computationally efficientapproach to detecting objects natively in 3D point clouds using convolutional neural networks (CNNs). In particular, this is achieved by leveraging a feature-centric voting scheme to implement novel convolutional layers which explicitly exploit the sparsity encountered in the input. To this end, we examine the trade-off between accuracy and speed for different architectures and additionally propose to use an L1 penalty on the filter activations to further encourage sparsity in the intermediate representations. To the best of our knowledge, this is the first work to propose sparse convolutional layers and L1 regularisation for efficient large-scale processing of 3D data.We demonstrate the efficacy of our approach on the KITTI object detection benchmark and show that Vote3Deep models with as few as three layers outperform the previous state of the art in both laser and laser-vision based approaches by margins of up to 40% while remaining highly competitive in terms of processing time.\n\n- i have downloaded the paper,interested readers can read full paper from here : https://github.com/mobassir94/ML-DL-Research-paper-Collection/blob/master/3D%20Deep%20Learning/vote3deep.pdf\n\n- Reference video link : https://www.youtube.com/watch?v=WUOSmAfeXIw", "678354": "Hi . Interesting . But it wil be applicable for Lidar type of data right ? Like SECOND ?", "678514": "hello @phoenix9032 \ngood question,\nthe paper mentioned can be applicable with lidar type of data and that cnn can also be used for this kind of competition problem", "680844": "Thank you for sharing! this is very interesting!", "680845": "glad it helps" }, "source": "meta" }