As part of Machine Perception course, I implemented the approach mentioned in Tomasi-Kanade Factorization method to extract shape and motion parameters from a sequence of images.
To run the code use the command python run.py FOLDER, where FOLDER is either medusa or castlejpg. This will generate a pointcloud text file with the 3d coordinates of the points and save each frame with detected keypoints. (I use SIFT Features)