Spaces:
Running
Running
Commit
•
664a6cd
1
Parent(s):
e0c6f6e
Upload 5 files
Browse files- CustomModel/config.json +27 -0
- CustomModel/model.safetensors +3 -0
- CustomModel/training_args.bin +3 -0
- app.py +109 -0
- requirements.txt +4 -0
CustomModel/config.json
ADDED
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "bert-base-uncased",
|
3 |
+
"architectures": [
|
4 |
+
"BertForSequenceClassification"
|
5 |
+
],
|
6 |
+
"attention_probs_dropout_prob": 0.1,
|
7 |
+
"classifier_dropout": null,
|
8 |
+
"gradient_checkpointing": false,
|
9 |
+
"hidden_act": "gelu",
|
10 |
+
"hidden_dropout_prob": 0.1,
|
11 |
+
"hidden_size": 768,
|
12 |
+
"initializer_range": 0.02,
|
13 |
+
"intermediate_size": 3072,
|
14 |
+
"layer_norm_eps": 1e-12,
|
15 |
+
"max_position_embeddings": 512,
|
16 |
+
"model_type": "bert",
|
17 |
+
"num_attention_heads": 12,
|
18 |
+
"num_hidden_layers": 12,
|
19 |
+
"pad_token_id": 0,
|
20 |
+
"position_embedding_type": "absolute",
|
21 |
+
"problem_type": "single_label_classification",
|
22 |
+
"torch_dtype": "float32",
|
23 |
+
"transformers_version": "4.39.3",
|
24 |
+
"type_vocab_size": 2,
|
25 |
+
"use_cache": true,
|
26 |
+
"vocab_size": 30522
|
27 |
+
}
|
CustomModel/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c0ae3b4736071ebf406209d00d51c502108761fafa3c8df37f6a009f0decb157
|
3 |
+
size 437958648
|
CustomModel/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:62f276a3fac2555bc29c7da8ad3095096c7ee3452711ca0c0cab720c0e053210
|
3 |
+
size 4920
|
app.py
ADDED
@@ -0,0 +1,109 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#run the app
|
2 |
+
#python -m streamlit run d:/NSFW/Project/test1.py
|
3 |
+
import torch
|
4 |
+
from transformers import BertTokenizer, BertForSequenceClassification
|
5 |
+
import math, keras_ocr
|
6 |
+
# Initialize pipeline
|
7 |
+
pipeline = keras_ocr.pipeline.Pipeline()
|
8 |
+
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
|
9 |
+
model_2 = BertForSequenceClassification.from_pretrained("CustomModel")
|
10 |
+
|
11 |
+
model_2.to('cpu')
|
12 |
+
import streamlit as st
|
13 |
+
|
14 |
+
def get_distance(predictions):
|
15 |
+
"""
|
16 |
+
Function returns dictionary with (key,value):
|
17 |
+
* text : detected text in image
|
18 |
+
* center_x : center of bounding box (x)
|
19 |
+
* center_y : center of bounding box (y)
|
20 |
+
* distance_from_origin : hypotenuse
|
21 |
+
* distance_y : distance between y and origin (0,0)
|
22 |
+
"""
|
23 |
+
|
24 |
+
# Point of origin
|
25 |
+
x0, y0 = 0, 0
|
26 |
+
|
27 |
+
# Generate dictionary
|
28 |
+
detections = []
|
29 |
+
for group in predictions:
|
30 |
+
|
31 |
+
# Get center point of bounding box
|
32 |
+
top_left_x, top_left_y = group[1][0]
|
33 |
+
bottom_right_x, bottom_right_y = group[1][1]
|
34 |
+
center_x, center_y = (top_left_x + bottom_right_x)/2, (top_left_y + bottom_right_y)/2
|
35 |
+
|
36 |
+
# Use the Pythagorean Theorem to solve for distance from origin
|
37 |
+
distance_from_origin = math.dist([x0,y0], [center_x, center_y])
|
38 |
+
|
39 |
+
# Calculate difference between y and origin to get unique rows
|
40 |
+
distance_y = center_y - y0
|
41 |
+
|
42 |
+
# Append all results
|
43 |
+
detections.append({
|
44 |
+
'text': group[0],
|
45 |
+
'center_x': center_x,
|
46 |
+
'center_y': center_y,
|
47 |
+
'distance_from_origin': distance_from_origin,
|
48 |
+
'distance_y': distance_y
|
49 |
+
})
|
50 |
+
|
51 |
+
return detections
|
52 |
+
|
53 |
+
def distinguish_rows(lst, thresh=15):
|
54 |
+
"""Function to help distinguish unique rows"""
|
55 |
+
sublists = []
|
56 |
+
for i in range(0, len(lst)-1):
|
57 |
+
if (lst[i+1]['distance_y'] - lst[i]['distance_y'] <= thresh):
|
58 |
+
if lst[i] not in sublists:
|
59 |
+
sublists.append(lst[i])
|
60 |
+
sublists.append(lst[i+1])
|
61 |
+
else:
|
62 |
+
yield sublists
|
63 |
+
sublists = [lst[i+1]]
|
64 |
+
yield sublists
|
65 |
+
|
66 |
+
# Title of the app
|
67 |
+
st.title("Image Input App")
|
68 |
+
|
69 |
+
# File uploader widget
|
70 |
+
uploaded_file = st.file_uploader("Upload an image", type=["jpg", "png", "jpeg"])
|
71 |
+
|
72 |
+
if uploaded_file is not None:
|
73 |
+
|
74 |
+
# Read in image
|
75 |
+
read_image = keras_ocr.tools.read(uploaded_file)
|
76 |
+
|
77 |
+
# prediction_groups is a list of (word, box) tuples
|
78 |
+
prediction_groups = pipeline.recognize([read_image])
|
79 |
+
predictions = prediction_groups[0] # extract text list
|
80 |
+
predictions = get_distance(predictions)
|
81 |
+
# Set thresh higher for text further apart
|
82 |
+
predictions = list(distinguish_rows(predictions, thresh=10))
|
83 |
+
|
84 |
+
# Remove all empty rows
|
85 |
+
predictions = list(filter(lambda x:x!=[], predictions))
|
86 |
+
|
87 |
+
# Order text detections in human readable format
|
88 |
+
ordered_preds = []
|
89 |
+
for row in predictions:
|
90 |
+
row = sorted(row, key=lambda x:x['distance_from_origin'])
|
91 |
+
for each in row: ordered_preds.append(each['text'])
|
92 |
+
|
93 |
+
# Join detections into sentence
|
94 |
+
sentance = ' '.join(ordered_preds)
|
95 |
+
#st.write(sentance)
|
96 |
+
|
97 |
+
text =sentance
|
98 |
+
print(text)
|
99 |
+
inputs = tokenizer(text,padding = True, truncation = True, return_tensors='pt').to('cpu')
|
100 |
+
outputs = model_2(**inputs)
|
101 |
+
predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
|
102 |
+
predictions = predictions.cpu().detach().numpy()
|
103 |
+
print(predictions[0][0],predictions[0][1])
|
104 |
+
if predictions[0][0]>predictions[0][1]:
|
105 |
+
print('safe')
|
106 |
+
st.write('safe')
|
107 |
+
else:
|
108 |
+
print('Not safe')
|
109 |
+
st.write('n safe')
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
transformers
|
3 |
+
keras_ocr
|
4 |
+
streamlit
|