Object Detection using Deep Neural Networks
- 1. 1 Object Detection By Usman Qayyum 4, Dec, 2018
- 2. Talk Covers Three Papers (Object Detection -> Embedded Computing) 2 SqueezeNet-2016SSD-2016 TinySSD-2018 =+
- 3. Image Classification/Object Detection ● Autonomous vehicles, smart video surveillance, facial detection and various applications, fast and robust object detection is need of an hour ● Nonly recognizing and classifying every object in an image, but localizing each one by drawing the appropriate bounding box around it. 3
- 4. CNN Migration (Image Classification) 4
- 5. Object Detection as Classification CNN deer? cat? background?
- 6. Object Detection as Classification CNN deer? cat? background?
- 7. Object Detection as Classification CNN deer? cat? background?
- 8. Object Detection as Classification with Sliding Window CNN deer? cat? background?
- 9. Object Detection as Classification with Box Proposals
- 10. Box Proposal Method : Selective Search Segmentation As Selective Search for Object Recognition. van de Sande et al. ICCV 2011
- 11. Idea behind Object Detectors ● Box Proposals ● Classifier Algorithm 11
- 12. RCNN Rich feature hierarchies for accurate object detection and semantic segmentation. Girshick et al. CVPR 2014. https://people.eecs.berkeley.edu/~rbg/papers/r-cnn-cvpr.pdf
- 13. Fast-RCNN Fast R-CNN. Girshick. ICCV 2015. https://arxiv.org/abs/1504.08083 Idea: No need to recompute features for every box independently, Regress refined bounding box coordinates.
- 14. Faster-RCNN Ren et al. NIPS 2015. https://arxiv.org/abs/1506.01497 Idea: Integrate the Bounding Box Propos als as part of the CNN predictions
- 15. YOLO- You Only Look Once ● Single Shot Detector Redmon et al. CVPR 2016. https://arxiv.org/abs/1506.02640 Idea: No bounding box proposals. Predict a class and a box for every location in a grid.
- 16. SSD: Single Shot Detector Liu et al. ECCV 2016. Idea: Similar to YOLO, but denser grid map, multiscale grid maps. + Data augm entation + Hard negative mining + Other design choices in the network.
- 17. -The overall objective loss function is a weighted sum of the localization loss and the confidence loss(conf) N: the number of matched default boxes l: predicted boxes g: the ground truth box x=1 denotes some certain default box is matched to a ground truth box17 1 ( , , , ) ( ( , ) ( , , ))conf locL x c l g L x c L x l g N SSD: Single Shot Detector
- 18. Performance 18
- 20. AI Workload Migration Embedded (Mobile/Edge) Server/Clou d Execution/Inference Training Execution/Inference Intelligence & Analytics Key Use Cases Vision | Audio | Security Benefits Low Latency | Privacy
- 21. AI in Embedded Devices 21
- 22. How ? (AI in Embedded Devices) Pruning Quantization22
- 23. SqueezeNet (Parameter Reduction) ● Strategy 1. Replace 3x3 filters with 1x1 filters ○ Parameters per filter: (3x3 filter) = 9 * (1x1 filter) ● Strategy 2. Decrease the number of input channels to 3x3 filters ○ Total # of parameters: (# of input channels) * (# of filters) * ( # of parameters per filter) ● Strategy 3. Downsample late in the network so that convolution layers have large activation maps ○ Size of activation maps: the size of input data, the choice of layers in which to downsample in the CNN architecture 23 Iandola, Forrest N., et al. "SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and< 0.5 MB model size."
- 24. Strategy#1 Conv1x1 or Kernel Reduction 24
- 25. Microarchitecture – Fire Module 25 Squeeze Layer Set s1x1 < (e1x1 + e3x3), limits the # of input channels to 3*3 filters Strategy 2. Decrease the number of input channels to 3x3 filters Total # of parameters: (# of input channels) * (# of filters) * ( # of parameters per filter) How much can we limit s1x1? Strategy 1. Replace 3*3 filters with 1*1 filters Parameters per filter: (3*3 filter) = 9 * (1*1 filter) How much can we replace 3*3 with 1*1? (e1x1 vs e3x3 )?
- 26. Expand ● In the "expand" modules, what are the tradeoffs when we turn the knob between mostly 1x1 and mostly 3x3 filters? ● Hypothesis: if having more weights leads to higher accuracy, then having all 3x3 filters should give the highest accuracy 27
- 27. 28
- 28. Macroarchitecture 29 Strategy 3. Downsample late in the network so that convolution layers have large activation maps Size of activation maps: the size of input data, the choice of layers in which to downsample in the CNN architecture
- 29. Performance 30
- 30. TinySSD (SSD with Microarchitecture) 31
- 31. Thanks for your attention. 32