Closed yxoh closed 2 years ago
Hi, you can download from the URLs
Like to follow this issue again...
I am wondering what's the best tool to convert urls to images? I have downloaded the "ccs_synthetic_filtered_large.json" linked from this repo, while using requests.get() to download each url roughly resulted in 15% invalid urls across first 10k files I downloaded.
Is this normal? Or I should follow the package linked from @woctezuma , but if using that method, it would not easy to apply the filtered captions generated from BLIP right?
Really appreciate for the help and detailed instructions..
can you share the URL to download the NoCaps dataset or the code to download nocap images? The official website only provides the URL for each image.
I also want to know how to get NoCaps dataset
can you share the URL to download the NoCaps dataset or the code to download nocap images? The official website only provides the URL for each image.