def main():
"""Visual Qustion Answering
Using Machine Learning and Streamlit
"""
__C = Cfgs()
args = parse_args()
args_dict = __C.parse_to_dict(args)
cfg_file = "cfgs/{}_model.yml".format(args.MODEL)
with open(cfg_file, 'r') as f:
yaml_dict = yaml.load(f)
args_dict = {**yaml_dict, **args_dict}
__C.add_args(args_dict)
__C.proc()
print('Hyper Parameters:')
print(__C)
st.title("VQA tool using Streamlit")
html_temp = """
<div style="background-color:tomato;padding:10px">
<h2 style="color:white;text-align:center;">ViPyKube ML App </h2>
</div>
"""
st.markdown(html_temp, unsafe_allow_html=True)
images = []
st.sidebar.title("Image selection and Question")
button = st.sidebar.radio(
'Randomly generate images',
('With predefined questons', 'With custom question'))
#while True:
img_path_list = []
for x in range(0, 10):
name = random.choice(os.listdir('./datasets/coco_extract/images'))
print(str(name))
name = './datasets/coco_extract/images/' + name
image = Image.open(name)
img_path_list.append(name)
images.append(image)
image_iterator = paginator("Select a sunset page", images)
indices_on_page, images_on_page = map(list, zip(*image_iterator))
st.image(images_on_page, width=200, caption=indices_on_page)
pick_img = st.sidebar.selectbox("Which image?",
[x for x in range(1, len(images))])
imp_path = img_path_list[int(pick_img)]
print("chosen image ", imp_path)
image = Image.open(imp_path)
st.header("Selected Image")
st.image(image)
img_id = imp_path[44:-4]
for ix in range(len(img_id)):
if img_id[ix] != "0":
img_id = img_id[ix:]
print(img_id)
break
q_list = []
if button == "With predefined questons":
pass
elif button == "With custom question":
question = st.sidebar.text_input("What is your question?")
q_list = [{
"image_id": int(img_id),
"question": question,
"question_id": 1
}]
start_eval = st.sidebar.button('Get the answer!')
if start_eval:
print('Loading testing set ........')
dataset = DataSet(__C, q_list, imp_path[31:], img_id)
eval(__C, dataset, valid=True)
parser.add_argument('--DATA_PATH',
dest='DATASET_PATH',
help='vqav2 dataset root path',
type=str)
parser.add_argument('--FEAT_PATH',
dest='FEATURE_PATH',
help='bottom up features root path',
type=str)
args = parser.parse_args()
return args
if __name__ == '__main__':
__C = Cfgs()
args = parse_args()
args_dict = __C.parse_to_dict(args)
cfg_file = "cfgs/{}_model.yml".format(args.MODEL)
with open(cfg_file, 'r') as f:
yaml_dict = yaml.load(f)
args_dict = {**yaml_dict, **args_dict}
__C.add_args(args_dict)
__C.proc()
print('Hyper Parameters:')
print(__C)
def json_store(path, file):
load_f = json.load(open(path, 'r', encoding='utf-8'))
with open(path, 'w') as f:
merge_file = {**load_f, **file}
f.write(json.dumps(merge_file, indent=4))
def init_json(path_list):
init = {}
for path in path_list:
with open(path, 'w') as f:
f.write(json.dumps(init, indent=4))
__C = Cfgs()
split_list = ['val', 'test']
stat_ques_list = []
path_list = ['datasets/vqa2db/ques_entities.json', 'datasets/vqa2db/failed_ques.json']
# init_json(path_list)
for i in split_list:
stat_ques_list += json.load(open(__C.QUESTION_PATH[i], 'r'))['questions']
# stat_ques_list = \
# json.load(open(__C.QUESTION_PATH['train'], 'r'))['questions'] + \
# json.load(open(__C.QUESTION_PATH['val'], 'r'))['questions'] + \
# json.load(open(__C.QUESTION_PATH['test'], 'r'))['questions'] + \
# json.load(open(__C.QUESTION_PATH['vg'], 'r'))['questions']