Computer vision
Budget: $30 – $250 USD
For each label we select 100 images to make a total of 10x100=1000 images. Objectives are using the gradient image features (SIFT, Dense SIFT, and HOG) with aggregation (VLAD, FisherVector) to test which one can give us the best solution.
[1] Gradient Features, for each image, implement the following function
[2] Compute the PCA and GMM model for each feature space
[3] Compute the gradient feature aggregations and benchmark its recognition accuracy
4 Compute the gradient feature VLAD aggregations and benchmark its recognition accuracy
[1] Gradient Features, for each image, implement the following function
[2] Compute the PCA and GMM model for each feature space
[3] Compute the gradient feature aggregations and benchmark its recognition accuracy
4 Compute the gradient feature VLAD aggregations and benchmark its recognition accuracy