Possible Bugs, Loader.py not working?

[SinghJasdeep]

Hello we are using you code to preprocess the data and build a few models on top of and are running into some bugs. Don't know why this is not working for us.

https://github.com/SinghJasdeep/CS230-CS224N-VQA

Mainly issues in the loader.py file:

i_batch.append(self.i_feat[iid]) is actually a dictionary not an array of size (36,2048)?

Your help would be greatly appreciated Mark.

[SinghJasdeep]

I think the error is possibly how we are saving coco_features.npy

Could you please post the version of read_tsv.py you use.

We are using this:

import base64
import numpy as np
import csv
import sys
import zlib
import time
import mmap

csv.field_size_limit(sys.maxsize)

FIELDNAMES = ['image_id', 'image_w', 'image_h', 'num_boxes', 'boxes', 'features']
infile = '/home/vhying/Downloads/CS230-CS224N-VQA/data/trainval_36/trainval_resnet101_faster_rcnn_genome_36.tsv'
outfile = '/home/vhying/Downloads/CS230-CS224N-VQA/data/coco_features.npy'

if name == 'main':

# Verify we can read a tsv
in_data = {}
with open(infile, "r+b") as tsv_in_file:
    reader = csv.DictReader(tsv_in_file, delimiter='\t', fieldnames=FIELDNAMES)
    for item in reader:
        item['image_id'] = int(item['image_id'])
        item['image_h'] = int(item['image_h'])
        item['image_w'] = int(item['image_w'])
        item['num_boxes'] = int(item['num_boxes'])
        for field in ['boxes', 'features']:
            item[field] = np.frombuffer(base64.decodestring(item[field]),
                                        dtype=np.float32).reshape((item['num_boxes'], -1))
        in_data[item['image_id']] = item

    with open(outfile, "wb") as Out_f:
        np.save(outfile, in_data)

[SinghJasdeep]

Issue Solved

[kevgeo]

Hi @SinghJasdeep , even I am facing a similar problem. How did you resolve the issue? Could you share your version of read_tsv.py? That would be great.

[markdtw]

Hi, sorry I really should've put my read_tsv.py here

it's like this

...
    in_data = {}
...
        for item in reader:
            data = {}
            data['image_id'] = int(item['image_id'])
            for field in ['features']:
                data[field] = np.frombuffer(base64.decodestring(item[field]),
                    dtype=np.float32).reshape((36, -1))
            in_data[data['image_id']] = data['features']

        np.save('purefeat.npy', in_data)
...

[kevgeo]

@markdtw Thank you so much for the code :).