HoG Descriptors and Bag of Words
Hi all, I'm new to computer vision/machine learning and I was hoping I could ask the community for some advise. I've calculated HoG descriptors for frames in a video but I'm not sure how best to group/join/??? them so I can then run Kmeans clustering on them. I'm hoping to use the (Visual) Bag of Words method to classify using random forrests but I'm a novice when it comes to ndarrays and not sure of the correct terminology. I know the HoG descriptors are flattened arrays but in order to cluster the frames/image descriptors I would need to group all the descriptors together. What is the best way to create a data structure suitable for kmeans when you have 100,000's of individual descriptors and do I need to pre-process the ndarrays ? Michael
Some properties of the individual ndarray (aka HoG descriptor) ('NDarry Size', 251328) ('NDarray Number of Dimensions', 1) ('NDarray length of 1 array element in Bytes', 8) ('NDarray Total byes consumed by elements', 2010624) ('NDarray DataType', dtype('float64')) On Sunday, 24 July 2016 14:59:04 UTC+1, Michael O'Brien wrote:
Hi all,
I'm new to computer vision/machine learning and I was hoping I could ask the community for some advise. I've calculated HoG descriptors for frames in a video but I'm not sure how best to group/join/??? them so I can then run Kmeans clustering on them. I'm hoping to use the (Visual) Bag of Words method to classify using random forrests but I'm a novice when it comes to ndarrays and not sure of the correct terminology.
I know the HoG descriptors are flattened arrays but in order to cluster the frames/image descriptors I would need to group all the descriptors together. What is the best way to create a data structure suitable for kmeans when you have 100,000's of individual descriptors and do I need to pre-process the ndarrays ?
Michael
And is there a way to select the top N HoG Descriptors from the flattened array? On Sunday, 24 July 2016 14:59:04 UTC+1, Michael O'Brien wrote:
Hi all,
I'm new to computer vision/machine learning and I was hoping I could ask the community for some advise. I've calculated HoG descriptors for frames in a video but I'm not sure how best to group/join/??? them so I can then run Kmeans clustering on them. I'm hoping to use the (Visual) Bag of Words method to classify using random forrests but I'm a novice when it comes to ndarrays and not sure of the correct terminology.
I know the HoG descriptors are flattened arrays but in order to cluster the frames/image descriptors I would need to group all the descriptors together. What is the best way to create a data structure suitable for kmeans when you have 100,000's of individual descriptors and do I need to pre-process the ndarrays ?
Michael
participants (1)
-
Michael O'Brien