Research Project

https://unix.stackexchange.com/questions/101237/how-to-recover-files-i-deleted-now-by-running-rm/101297

define the problem:    (1) single class(pedestrian) or multi-class?   single(pedestrian)    (2) in Mask-RCNN(Faster-RCNN), NMS be applied in both training and inference, our target just replaces the inference part or both parts?    inference    (3) at the moment, are we mainly focus on the single image or video frames?    single image    (4) are we just focus on detection or both detection and segmentation?  detection, but need to use segmentation to help detection.    (5) shall we use tight_boxes or lose_boxes as gt_annots?  (which make difference for gt_masks) related papers: cat1: pairwise(clustering) 1. Non-maximum suppression for object detection by passing messages between windows_Rasmus Rothe_ACCV2014 https://www.vision.ee.ethz.ch/publications/papers/proceedings/eth_biwi_01126.pdf based on Affinity propagation clustering(APC), using a Latent Structured SVM(LSSVM) to learn the weigh...

1.softmax losss and cross-entropy http://cs231n.github.io/linear-classify/ 2. sigmoid: https://en.wikipedia.org/wiki/Sigmoid_function

1.  tud-crossing [1] H. Riemenschneider, S. Sternig. M. Donoser, P. Roth and H. Bischof.     Hough Regions for Joining Instance Localization and Segmentation. ECCV 2012 Contents: tud-crossing-sequence_segmentation_ICG_anno.txt - new bbox annotation DaSide0811-seq7-*_bbox.png                      - new bbox overlayed in color DaSide0811-seq7-*_bunt.png                      - new segmentation in color DaSide0811-seq7-*_over.png                      - new segmentation overlayed to image DaSide0811-seq7-*_seg.png                       - new segmentation as labels The format of each annotation line is: frame,x1,y1,x2,y2,visibility, ASC(l),ASC(r),ASC(h),ASC(f),ASC(b), uniqueID Final notes:  The segmentation masks were individually and ma...

1. 2. coco prototools https://github.com/waleedka/coco coco tools ~/uniLab/projects/tools$  git clone --recursive https://github.com/waleedka/coco.git 2.1 enter into  tools/coco/PythonAPI edit makeFile: replace python with python3 (for python3.n) 2.2 $ make 2.4. create a symbolink: /uniLab/projects/Python/Mask_RCNN$  ln -s /home/uni/uniLab/projects/tools/coco/PythonAPI/pycocotools  pycocotools 3. prepare data  /Mask_RCNN$ cd data /Mask_RCNN/data$ ln -s /home/uni/uniLab/dataset/coco coco  

bug: error while loading shared libraries: libldof.so: cannot open shared object file: No such file or directory. solution:(sudo permission) https://gist.github.com/simplay/e9752771d1c607b35fab

user:  ssh -X uni@robotvision1.cs.adelaide.edu.au pass: same as local machine admin: @robotvision1.cs.adelaide.edu.au install python3.5.2 https://tecadmin.net/install-python-3-5-on-ubuntu/