tosolveit / triplet Goto Github PK

For image size 299x299x3 reading images from disk and obtaining embedding by running through network and writing embedding to the index takes 3.75 sec. per image, this slows down model evaluation and needs some improvement.

Some suggestions:

• Do analysis and identify the exact step or steps that slows down the processes
• Add progress bar for with tqdm for evaluate.py/created_index
• Multi-core approaches can also be used for speed up