Skosana, Vusi JNgxande, M2021-03-032021-03-032020-01Skosana, V.J. & Ngxande, M. 2020. The effects of image smoothing on CNN-based detectors. http://hdl.handle.net/10204/11821 .978-1-7281-4162-6978-1-7281-4163-3http://hdl.handle.net/10204/11821Edge-preserving smoothing filters have been shown to improve generalisation performance on the HOG features with a SVM classifier. However, not all smoothing filters and parameters lead to better performance. The effects of smoothing filters are studied on the Faster R-CNN detector using generic object and human detection datasets, namely the PASCAL VOC and KITTI respectively. The total variation (TV) smoothing filter was used for this study. It was found that the TV smoothing removed details the CNN was using for detection which degraded performance for both datasets. The results are consistent with previous observations that CNNs tend to learn weak visual features. The performance loss, however, was moderate and could be justified in the context of improving robustness to perturbations. The PASCAL VOC and KITTI datasets had comparable performance loss despite the latter having many more small objects that tend to blend into the background when smoothing is applied.AbstractenObject detectionImage smoothingEdge-preserving filterHuman detectionPedestrian detectionConvolutional neural networksConference PresentationSkosana, V. J., & Ngxande, M. (2020). The effects of image smoothing on CNN-based detectors. http://hdl.handle.net/10204/11821Skosana, Vusi J, and M Ngxande. "The effects of image smoothing on CNN-based detectors." <i>2020 International SAUPEC/RobMech/PRASA Conference, Cape Town, South Africa</i> (2020): http://hdl.handle.net/10204/11821Skosana VJ, Ngxande M, The effects of image smoothing on CNN-based detectors; 2020. http://hdl.handle.net/10204/11821 .TY - Conference Presentation AU - Skosana, Vusi J AU - Ngxande, M AB - Edge-preserving smoothing filters have been shown to improve generalisation performance on the HOG features with a SVM classifier. However, not all smoothing filters and parameters lead to better performance. The effects of smoothing filters are studied on the Faster R-CNN detector using generic object and human detection datasets, namely the PASCAL VOC and KITTI respectively. The total variation (TV) smoothing filter was used for this study. It was found that the TV smoothing removed details the CNN was using for detection which degraded performance for both datasets. The results are consistent with previous observations that CNNs tend to learn weak visual features. The performance loss, however, was moderate and could be justified in the context of improving robustness to perturbations. The PASCAL VOC and KITTI datasets had comparable performance loss despite the latter having many more small objects that tend to blend into the background when smoothing is applied. DA - 2020-01 DB - ResearchSpace DP - CSIR J1 - 2020 International SAUPEC/RobMech/PRASA Conference, Cape Town, South Africa KW - Object detection KW - Image smoothing KW - Edge-preserving filter KW - Human detection KW - Pedestrian detection KW - Convolutional neural networks LK - https://researchspace.csir.co.za PY - 2020 SM - 978-1-7281-4162-6 SM - 978-1-7281-4163-3 T1 - The effects of image smoothing on CNN-based detectors TI - The effects of image smoothing on CNN-based detectors UR - http://hdl.handle.net/10204/11821 ER -24284