Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -10,151 +10,140 @@ from my_model.object_detection import detect_and_draw_objects
|
|
10 |
from my_model.captioner.image_captioning import get_caption
|
11 |
from my_model.utilities import free_gpu_resources
|
12 |
from my_model.KBVQA import KBVQA, prepare_kbvqa_model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
|
|
|
|
|
|
|
|
|
14 |
|
15 |
-
|
16 |
-
def answer_question(image, question, caption, detected_objects_str, model):
|
17 |
-
|
18 |
-
answer = model.generate_answer(question, caption, detected_objects_str)
|
19 |
-
st.image(image)
|
20 |
-
st.write(caption)
|
21 |
-
st.write("----------------")
|
22 |
-
st.write(detected_objects_str)
|
23 |
-
return answer
|
24 |
-
|
25 |
-
def get_caption(image):
|
26 |
-
return "Generated caption for the image"
|
27 |
-
|
28 |
-
def free_gpu_resources():
|
29 |
-
pass
|
30 |
-
|
31 |
-
# Sample images (assuming these are paths to your sample images)
|
32 |
-
sample_images = ["Files/sample1.jpg", "Files/sample2.jpg", "Files/sample3.jpg",
|
33 |
-
"Files/sample4.jpg", "Files/sample5.jpg", "Files/sample6.jpg",
|
34 |
-
"Files/sample7.jpg"]
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
def analyze_image(image, model, show_processed_image=False):
|
39 |
-
img = copy.deepcopy(image)
|
40 |
-
caption = model.get_caption(img)
|
41 |
-
image_with_boxes, detected_objects_str = model.detect_objects(img)
|
42 |
-
if show_processed_image:
|
43 |
-
st.image(image_with_boxes)
|
44 |
-
return caption, detected_objects_str
|
45 |
-
|
46 |
-
def image_qa_app(kbvqa):
|
47 |
-
# Initialize session state for storing the current image and its Q&A history
|
48 |
-
if 'current_image' not in st.session_state:
|
49 |
-
st.session_state['current_image'] = None
|
50 |
-
if 'qa_history' not in st.session_state:
|
51 |
-
st.session_state['qa_history'] = []
|
52 |
-
if 'analysis_done' not in st.session_state:
|
53 |
-
st.session_state['analysis_done'] = False
|
54 |
-
if 'answer_in_progress' not in st.session_state:
|
55 |
-
st.session_state['answer_in_progress'] = False
|
56 |
-
|
57 |
-
# Display sample images as clickable thumbnails
|
58 |
-
st.write("Choose from sample images:")
|
59 |
-
cols = st.columns(len(sample_images))
|
60 |
-
for idx, sample_image_path in enumerate(sample_images):
|
61 |
-
with cols[idx]:
|
62 |
-
image = Image.open(sample_image_path)
|
63 |
-
st.image(image, use_column_width=True)
|
64 |
-
if st.button(f'Select Sample Image {idx + 1}', key=f'sample_{idx}'):
|
65 |
-
st.session_state['current_image'] = image
|
66 |
-
st.session_state['qa_history'] = []
|
67 |
-
st.session_state['analysis_done'] = False
|
68 |
-
st.session_state['answer_in_progress'] = False
|
69 |
-
|
70 |
-
# Image uploader
|
71 |
uploaded_image = st.file_uploader("Or upload an Image", type=["png", "jpg", "jpeg"])
|
72 |
if uploaded_image is not None:
|
73 |
-
|
74 |
-
st.session_state['current_image'] = image
|
75 |
st.session_state['qa_history'] = []
|
76 |
st.session_state['analysis_done'] = False
|
77 |
st.session_state['answer_in_progress'] = False
|
78 |
|
79 |
if st.session_state.get('current_image') and not st.session_state.get('analysis_done', False):
|
80 |
if st.button('Analyze Image'):
|
81 |
-
caption, detected_objects_str = analyze_image(st.session_state['current_image'],
|
82 |
st.session_state['caption'] = caption
|
83 |
st.session_state['detected_objects_str'] = detected_objects_str
|
84 |
st.session_state['analysis_done'] = True
|
85 |
|
86 |
-
# Get Answer button
|
87 |
if st.session_state.get('analysis_done', False):
|
88 |
question = st.text_input("Ask a question about this image:")
|
89 |
if st.button('Get Answer'):
|
90 |
-
answer = answer_question(
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
st.session_state['qa_history'].append((question, answer))
|
92 |
|
93 |
-
# Display all Q&A
|
94 |
for q, a in st.session_state.get('qa_history', []):
|
95 |
st.text(f"Q: {q}\nA: {a}\n")
|
96 |
|
97 |
-
# Reset the answer_in_progress flag after displaying the answer
|
98 |
-
if st.session_state['answer_in_progress']:
|
99 |
-
st.session_state['answer_in_progress'] = False
|
100 |
-
|
101 |
def run_inference():
|
|
|
102 |
st.title("Run Inference")
|
103 |
|
104 |
method = st.selectbox(
|
105 |
"Choose a method:",
|
106 |
["Fine-Tuned Model", "In-Context Learning (n-shots)"],
|
107 |
-
index=0
|
108 |
)
|
109 |
|
110 |
-
detection_model = st.selectbox(
|
111 |
-
"Choose a model for object detection:",
|
112 |
-
["yolov5", "detic"],
|
113 |
-
index=0 # Default to the first option
|
114 |
-
)
|
115 |
-
|
116 |
-
# Initialize session state for the model
|
117 |
if method == "Fine-Tuned Model":
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
st.write("Model is ready for inference.")
|
131 |
-
# Set default confidence based on the selected model
|
132 |
-
default_confidence = 0.2 if detection_model == "yolov5" else 0.4
|
133 |
-
# Slider for confidence level
|
134 |
-
confidence_level = st.slider(
|
135 |
-
"Select Detection Confidence Level",
|
136 |
-
min_value=0.1,
|
137 |
-
max_value=0.9,
|
138 |
-
value=default_confidence,
|
139 |
-
step=0.1
|
140 |
-
)
|
141 |
-
st.session_state['kbvqa'].detection_confidence = confidence_level
|
142 |
|
143 |
-
if st.session_state['kbvqa']:
|
144 |
-
image_qa_app(st.session_state['kbvqa'])
|
145 |
|
146 |
-
else:
|
147 |
-
st.write(f'{method} model is not ready for inference yet')
|
148 |
-
|
149 |
-
# Main function
|
150 |
def main():
|
151 |
st.sidebar.title("Navigation")
|
152 |
selection = st.sidebar.radio("Go to", ["Home", "Dataset Analysis", "Evaluation Results", "Run Inference", "Dissertation Report"])
|
153 |
|
154 |
if selection == "Home":
|
155 |
-
st.title("MultiModal Learning for
|
156 |
st.write("Home page content goes here...")
|
157 |
-
|
158 |
elif selection == "Dissertation Report":
|
159 |
st.title("Dissertation Report")
|
160 |
st.write("Click the link below to view the PDF.")
|
@@ -166,22 +155,29 @@ def main():
|
|
166 |
mime="application/octet-stream"
|
167 |
)
|
168 |
|
169 |
-
|
170 |
elif selection == "Evaluation Results":
|
171 |
st.title("Evaluation Results")
|
172 |
st.write("This is a Place Holder until the contents are uploaded.")
|
173 |
|
174 |
-
|
175 |
elif selection == "Dataset Analysis":
|
176 |
st.title("OK-VQA Dataset Analysis")
|
177 |
st.write("This is a Place Holder until the contents are uploaded.")
|
178 |
|
179 |
-
|
180 |
elif selection == "Run Inference":
|
181 |
run_inference()
|
182 |
-
|
183 |
-
|
184 |
-
|
|
|
|
|
|
|
185 |
|
186 |
if __name__ == "__main__":
|
187 |
-
main()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
from my_model.captioner.image_captioning import get_caption
|
11 |
from my_model.utilities import free_gpu_resources
|
12 |
from my_model.KBVQA import KBVQA, prepare_kbvqa_model
|
13 |
+
import my_model.utilities.st_config as st_config
|
14 |
+
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
class ImageHandler:
|
19 |
+
@staticmethod
|
20 |
+
def analyze_image(image, model, show_processed_image=False):
|
21 |
+
img = copy.deepcopy(image)
|
22 |
+
caption = model.get_caption(img)
|
23 |
+
image_with_boxes, detected_objects_str = model.detect_objects(img)
|
24 |
+
if show_processed_image:
|
25 |
+
st.image(image_with_boxes)
|
26 |
+
return caption, detected_objects_str
|
27 |
+
|
28 |
+
@staticmethod
|
29 |
+
def free_gpu_resources():
|
30 |
+
# Implementation for freeing GPU resources
|
31 |
+
free_gpu_resources()
|
32 |
+
|
33 |
+
class QuestionAnswering:
|
34 |
+
@staticmethod
|
35 |
+
def answer_question(image, question, caption, detected_objects_str, model):
|
36 |
+
answer = model.generate_answer(question, caption, detected_objects_str)
|
37 |
+
st.image(image)
|
38 |
+
st.write(caption)
|
39 |
+
st.write("----------------")
|
40 |
+
st.write(detected_objects_str)
|
41 |
+
return answer
|
42 |
+
|
43 |
+
class UIComponents:
|
44 |
+
@staticmethod
|
45 |
+
def display_image_selection(sample_images):
|
46 |
+
cols = st.columns(len(sample_images))
|
47 |
+
for idx, sample_image_path in enumerate(sample_images):
|
48 |
+
with cols[idx]:
|
49 |
+
image = Image.open(sample_image_path)
|
50 |
+
st.image(image, use_column_width=True)
|
51 |
+
if st.button(f'Select Sample Image {idx + 1}', key=f'sample_{idx}'):
|
52 |
+
st.session_state['current_image'] = image
|
53 |
+
st.session_state['qa_history'] = []
|
54 |
+
st.session_state['analysis_done'] = False
|
55 |
+
st.session_state['answer_in_progress'] = False
|
56 |
+
|
57 |
+
def load_kbvqa_model(detection_model):
|
58 |
+
"""Load KBVQA Model based on the selected detection model."""
|
59 |
+
if st.session_state.get('kbvqa') is not None:
|
60 |
+
st.write("Model already loaded.")
|
61 |
+
else:
|
62 |
+
st.session_state['kbvqa'] = prepare_kbvqa_model(detection_model)
|
63 |
+
if st.session_state['kbvqa']:
|
64 |
+
st.write("Model is ready for inference.")
|
65 |
+
return True
|
66 |
+
return False
|
67 |
+
|
68 |
+
def set_model_confidence(detection_model):
|
69 |
+
"""Set the confidence level for the detection model."""
|
70 |
+
default_confidence = 0.2 if detection_model == "yolov5" else 0.4
|
71 |
+
confidence_level = st.slider(
|
72 |
+
"Select Detection Confidence Level",
|
73 |
+
min_value=0.1,
|
74 |
+
max_value=0.9,
|
75 |
+
value=default_confidence,
|
76 |
+
step=0.1
|
77 |
+
)
|
78 |
+
st.session_state['kbvqa'].detection_confidence = confidence_level
|
79 |
|
80 |
+
def image_qa_app(kbvqa_model):
|
81 |
+
"""Streamlit app interface for image QA."""
|
82 |
+
sample_images = st_config.SAMPLE_IMAGES
|
83 |
+
UIComponents.display_image_selection(sample_images)
|
84 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
uploaded_image = st.file_uploader("Or upload an Image", type=["png", "jpg", "jpeg"])
|
86 |
if uploaded_image is not None:
|
87 |
+
st.session_state['current_image'] = Image.open(uploaded_image)
|
|
|
88 |
st.session_state['qa_history'] = []
|
89 |
st.session_state['analysis_done'] = False
|
90 |
st.session_state['answer_in_progress'] = False
|
91 |
|
92 |
if st.session_state.get('current_image') and not st.session_state.get('analysis_done', False):
|
93 |
if st.button('Analyze Image'):
|
94 |
+
caption, detected_objects_str = ImageHandler.analyze_image(st.session_state['current_image'], kbvqa_model)
|
95 |
st.session_state['caption'] = caption
|
96 |
st.session_state['detected_objects_str'] = detected_objects_str
|
97 |
st.session_state['analysis_done'] = True
|
98 |
|
|
|
99 |
if st.session_state.get('analysis_done', False):
|
100 |
question = st.text_input("Ask a question about this image:")
|
101 |
if st.button('Get Answer'):
|
102 |
+
answer = QuestionAnswering.answer_question(
|
103 |
+
st.session_state['current_image'],
|
104 |
+
question,
|
105 |
+
st.session_state.get('caption', ''),
|
106 |
+
st.session_state.get('detected_objects_str', ''),
|
107 |
+
kbvqa_model
|
108 |
+
)
|
109 |
st.session_state['qa_history'].append((question, answer))
|
110 |
|
|
|
111 |
for q, a in st.session_state.get('qa_history', []):
|
112 |
st.text(f"Q: {q}\nA: {a}\n")
|
113 |
|
|
|
|
|
|
|
|
|
114 |
def run_inference():
|
115 |
+
"""Main function to run inference based on the selected method."""
|
116 |
st.title("Run Inference")
|
117 |
|
118 |
method = st.selectbox(
|
119 |
"Choose a method:",
|
120 |
["Fine-Tuned Model", "In-Context Learning (n-shots)"],
|
121 |
+
index=0
|
122 |
)
|
123 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
124 |
if method == "Fine-Tuned Model":
|
125 |
+
detection_model = st.selectbox(
|
126 |
+
"Choose a model for object detection:",
|
127 |
+
["yolov5", "detic"],
|
128 |
+
index=0
|
129 |
+
)
|
130 |
+
|
131 |
+
if 'kbvqa' not in st.session_state or st.session_state['detection_model'] != detection_model:
|
132 |
+
st.session_state['detection_model'] = detection_model
|
133 |
+
if load_kbvqa_model(detection_model):
|
134 |
+
set_model_confidence(detection_model)
|
135 |
+
image_qa_app(st.session_state['kbvqa'])
|
136 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
|
|
|
|
|
138 |
|
|
|
|
|
|
|
|
|
139 |
def main():
|
140 |
st.sidebar.title("Navigation")
|
141 |
selection = st.sidebar.radio("Go to", ["Home", "Dataset Analysis", "Evaluation Results", "Run Inference", "Dissertation Report"])
|
142 |
|
143 |
if selection == "Home":
|
144 |
+
st.title("MultiModal Learning for Knowledge-Based Visual Question Answering")
|
145 |
st.write("Home page content goes here...")
|
146 |
+
|
147 |
elif selection == "Dissertation Report":
|
148 |
st.title("Dissertation Report")
|
149 |
st.write("Click the link below to view the PDF.")
|
|
|
155 |
mime="application/octet-stream"
|
156 |
)
|
157 |
|
|
|
158 |
elif selection == "Evaluation Results":
|
159 |
st.title("Evaluation Results")
|
160 |
st.write("This is a Place Holder until the contents are uploaded.")
|
161 |
|
|
|
162 |
elif selection == "Dataset Analysis":
|
163 |
st.title("OK-VQA Dataset Analysis")
|
164 |
st.write("This is a Place Holder until the contents are uploaded.")
|
165 |
|
|
|
166 |
elif selection == "Run Inference":
|
167 |
run_inference()
|
168 |
+
|
169 |
+
|
170 |
+
|
171 |
+
|
172 |
+
|
173 |
+
|
174 |
|
175 |
if __name__ == "__main__":
|
176 |
+
main()
|
177 |
+
|
178 |
+
|
179 |
+
|
180 |
+
|
181 |
+
|
182 |
+
|
183 |
+
|