the-full-stack / fsdl-text-recognizer-project

Lab materials for the Full Stack Deep Learning Course
https://fullstackdeeplearning.com/course
1.21k stars 428 forks source link

Dataset cant be accessed from S3 url #15

Open phospheneai opened 4 years ago

phospheneai commented 4 years ago

all data set resides in cloud cant be retrieved , 403 forbidden error

yunh-kwak commented 4 years ago

To solve the problem, Replace url at fsdl-text-recognizer-project/data/raw/emnist/metadata.toml with http://www.itl.nist.gov/iaui/vip/cs_links/EMNIST/matlab.zip

But It is EXTREMELY slow😂

sebastian-sosa commented 4 years ago

To avoid suffering those download speeds again, I backed up the byclass.h5 file generated under data/processed/mnist. You can download it from here, if it helps: https://mega.nz/#F!B6QFVQSI (decryption key 8ODxY82iUjLCbW3EUuiD0g)

Besides downloading the .mat file and creating the HDF5 file, the script also generates the JSON file lab2/text_recognizer/datasets/emnist_essentials.json, which I also uploaded to that folder.