PFNet Point Fractal Network for 3 D Point
![PF-Net: Point Fractal Network for 3 D Point Cloud Completion CVPR 2020 Zitian Huang PF-Net: Point Fractal Network for 3 D Point Cloud Completion CVPR 2020 Zitian Huang](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-1.jpg)
![Task: 3 D Point Cloud Completion Task: 3 D Point Cloud Completion](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-2.jpg)
![Compare with previous methods • genus-wise distortions • missing cross-bar input LGAN-AE PCN 3 Compare with previous methods • genus-wise distortions • missing cross-bar input LGAN-AE PCN 3](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-3.jpg)
![Method overview Multi-resolution encoder-decoder Method overview Multi-resolution encoder-decoder](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-4.jpg)
![IFPS downsampling • a method from POINTNet++ • iterative farthest point sampling We want IFPS downsampling • a method from POINTNet++ • iterative farthest point sampling We want](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-5.jpg)
![CMLP • Reserve all features from inner layers CMLP • Reserve all features from inner layers](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-6.jpg)
![Point Pyramid Decoder • FC and conv • Compute loss of three resolutions Point Pyramid Decoder • FC and conv • Compute loss of three resolutions](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-7.jpg)
![Adversarial Loss • Further boost performance Adversarial Loss • Further boost performance](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-8.jpg)
![Experiments Dataset • benchmark dataset Shapenet-Part • 13 categories are used • total number Experiments Dataset • benchmark dataset Shapenet-Part • 13 categories are used • total number](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-9.jpg)
![Experiments Metric • Pred → GT (prediction to ground truth) error • and GT Experiments Metric • Pred → GT (prediction to ground truth) error • and GT](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-10.jpg)
![Compare with SOTA Point cloud completion results of overall point cloud. Compare with SOTA Point cloud completion results of overall point cloud.](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-11.jpg)
![Compare with SOTA Point cloud completion results of the missing point cloud. Compare with SOTA Point cloud completion results of the missing point cloud.](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-12.jpg)
![](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-13.jpg)
![Robustness Test 1 • Lose different proportion Yellow represents the prediction. Grey denotes the Robustness Test 1 • Lose different proportion Yellow represents the prediction. Grey denotes the](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-14.jpg)
![Robustness Test 2 • loss points in two random positions each time Yellow represents Robustness Test 2 • loss points in two random positions each time Yellow represents](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-15.jpg)
![Comments + genus-wise distortions (propose to only predict missing part) + If we regard Comments + genus-wise distortions (propose to only predict missing part) + If we regard](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-16.jpg)
- Slides: 16
![PFNet Point Fractal Network for 3 D Point Cloud Completion CVPR 2020 Zitian Huang PF-Net: Point Fractal Network for 3 D Point Cloud Completion CVPR 2020 Zitian Huang](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-1.jpg)
PF-Net: Point Fractal Network for 3 D Point Cloud Completion CVPR 2020 Zitian Huang 1, Yikuan Yu 1, 2, Jiawen Xu 1, Feng Ni 2, Xinyi Le 1∗ 1 Shanghai Jiao Tong University, 2 Sense. Time Present by Zhixuan Li
![Task 3 D Point Cloud Completion Task: 3 D Point Cloud Completion](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-2.jpg)
Task: 3 D Point Cloud Completion
![Compare with previous methods genuswise distortions missing crossbar input LGANAE PCN 3 Compare with previous methods • genus-wise distortions • missing cross-bar input LGAN-AE PCN 3](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-3.jpg)
Compare with previous methods • genus-wise distortions • missing cross-bar input LGAN-AE PCN 3 D-Capsule This paper GT
![Method overview Multiresolution encoderdecoder Method overview Multi-resolution encoder-decoder](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-4.jpg)
Method overview Multi-resolution encoder-decoder
![IFPS downsampling a method from POINTNet iterative farthest point sampling We want IFPS downsampling • a method from POINTNet++ • iterative farthest point sampling We want](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-5.jpg)
IFPS downsampling • a method from POINTNet++ • iterative farthest point sampling We want to sample N points 1. randomly choose one initial point 2. find the farthest point, add to sampled set 3. iteratively execute for N-1 times Advantage: IFPS can represent the distribution of the entire point sets better compared to random sampling, and it is more efficient than CNNs * Point. Net++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space, NIPS 2017
![CMLP Reserve all features from inner layers CMLP • Reserve all features from inner layers](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-6.jpg)
CMLP • Reserve all features from inner layers
![Point Pyramid Decoder FC and conv Compute loss of three resolutions Point Pyramid Decoder • FC and conv • Compute loss of three resolutions](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-7.jpg)
Point Pyramid Decoder • FC and conv • Compute loss of three resolutions
![Adversarial Loss Further boost performance Adversarial Loss • Further boost performance](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-8.jpg)
Adversarial Loss • Further boost performance
![Experiments Dataset benchmark dataset ShapenetPart 13 categories are used total number Experiments Dataset • benchmark dataset Shapenet-Part • 13 categories are used • total number](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-9.jpg)
Experiments Dataset • benchmark dataset Shapenet-Part • 13 categories are used • total number of shapes sums to 14473 (11705 for training and 2768 for testing)
![Experiments Metric Pred GT prediction to ground truth error and GT Experiments Metric • Pred → GT (prediction to ground truth) error • and GT](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-10.jpg)
Experiments Metric • Pred → GT (prediction to ground truth) error • and GT → Pred (ground truth to prediction) error • Pred → GT error computes the average squared distance from each point in prediction to its closest in ground truth. • It measures how difference the prediction is from the ground truth.
![Compare with SOTA Point cloud completion results of overall point cloud Compare with SOTA Point cloud completion results of overall point cloud.](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-11.jpg)
Compare with SOTA Point cloud completion results of overall point cloud.
![Compare with SOTA Point cloud completion results of the missing point cloud Compare with SOTA Point cloud completion results of the missing point cloud.](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-12.jpg)
Compare with SOTA Point cloud completion results of the missing point cloud.
![](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-13.jpg)
![Robustness Test 1 Lose different proportion Yellow represents the prediction Grey denotes the Robustness Test 1 • Lose different proportion Yellow represents the prediction. Grey denotes the](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-14.jpg)
Robustness Test 1 • Lose different proportion Yellow represents the prediction. Grey denotes the undamaged point cloud. lose 25%, 75% points 50%,
![Robustness Test 2 loss points in two random positions each time Yellow represents Robustness Test 2 • loss points in two random positions each time Yellow represents](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-15.jpg)
Robustness Test 2 • loss points in two random positions each time Yellow represents the prediction. Grey denotes the undamaged point cloud.
![Comments genuswise distortions propose to only predict missing part If we regard Comments + genus-wise distortions (propose to only predict missing part) + If we regard](https://slidetodoc.com/presentation_image_h2/cb0c14265b00ab4fb5038c2972709797/image-16.jpg)
Comments + genus-wise distortions (propose to only predict missing part) + If we regard segmentation mask as outlier curve, may be the amodal completion problem could be transferred as this paper
What is the mountain
Fractals
Chaos theory explained
Cantor set fractal dimension
Fractal terrains 4
What is fractal art
What is a fractal
Helecho de barnsley
Triangle lmn has the vertices shown on the coordinate grid
What does this mean
Fractal image compression example
Fractal curve in computer graphics
Arte fractal
Space filling fractal
Fractal dimension definition
Transformasi tingkat keabuan
Chia compression