Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -13,9 +13,13 @@ from my_model.KBVQA import KBVQA, prepare_kbvqa_model
|
|
| 13 |
|
| 14 |
|
| 15 |
|
| 16 |
-
def answer_question(image, question, model):
|
| 17 |
|
| 18 |
-
answer = model.generate_answer(question, image)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
return answer
|
| 20 |
|
| 21 |
def get_caption(image):
|
|
@@ -31,14 +35,13 @@ sample_images = ["Files/sample1.jpg", "Files/sample2.jpg", "Files/sample3.jpg",
|
|
| 31 |
|
| 32 |
|
| 33 |
|
| 34 |
-
def analyze_image(image, model):
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
pass
|
| 42 |
|
| 43 |
def image_qa_app(kbvqa):
|
| 44 |
# Initialize session state for storing the current image and its Q&A history
|
|
@@ -77,7 +80,7 @@ def image_qa_app(kbvqa):
|
|
| 77 |
if st.session_state.get('current_image') and not st.session_state['analysis_done']:
|
| 78 |
if st.button('Analyze Image'):
|
| 79 |
# Perform analysis on the image
|
| 80 |
-
analyze_image(st.session_state['current_image'], kbvqa)
|
| 81 |
st.session_state['analysis_done'] = True
|
| 82 |
st.session_state['processed_image'] = copy.deepcopy(st.session_state['current_image'])
|
| 83 |
|
|
@@ -90,7 +93,7 @@ def image_qa_app(kbvqa):
|
|
| 90 |
question = st.text_input("Ask a question about this image:")
|
| 91 |
if st.button('Get Answer'):
|
| 92 |
st.session_state['answer_in_progress'] = True
|
| 93 |
-
answer = answer_question(st.session_state['processed_image'], question, model=kbvqa)
|
| 94 |
st.session_state['qa_history'].append((question, answer))
|
| 95 |
|
| 96 |
|
|
@@ -153,7 +156,7 @@ def run_inference():
|
|
| 153 |
# Main function
|
| 154 |
def main():
|
| 155 |
st.sidebar.title("Navigation")
|
| 156 |
-
selection = st.sidebar.radio("Go to", ["Home", "Dataset Analysis", "Evaluation Results", "Run Inference", "Dissertation Report"
|
| 157 |
|
| 158 |
if selection == "Home":
|
| 159 |
st.title("MultiModal Learning for Knowledg-Based Visual Question Answering")
|
|
|
|
| 13 |
|
| 14 |
|
| 15 |
|
| 16 |
+
def answer_question(image, question, caption, detected_objects_str, model):
|
| 17 |
|
| 18 |
+
answer = model.generate_answer(question, image, 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):
|
|
|
|
| 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
|
|
|
|
| 45 |
|
| 46 |
def image_qa_app(kbvqa):
|
| 47 |
# Initialize session state for storing the current image and its Q&A history
|
|
|
|
| 80 |
if st.session_state.get('current_image') and not st.session_state['analysis_done']:
|
| 81 |
if st.button('Analyze Image'):
|
| 82 |
# Perform analysis on the image
|
| 83 |
+
caption, detected_objects = analyze_image(st.session_state['current_image'], kbvqa)
|
| 84 |
st.session_state['analysis_done'] = True
|
| 85 |
st.session_state['processed_image'] = copy.deepcopy(st.session_state['current_image'])
|
| 86 |
|
|
|
|
| 93 |
question = st.text_input("Ask a question about this image:")
|
| 94 |
if st.button('Get Answer'):
|
| 95 |
st.session_state['answer_in_progress'] = True
|
| 96 |
+
answer = answer_question(st.session_state['processed_image'], question, caption, detected_objects_str, model=kbvqa)
|
| 97 |
st.session_state['qa_history'].append((question, answer))
|
| 98 |
|
| 99 |
|
|
|
|
| 156 |
# Main function
|
| 157 |
def main():
|
| 158 |
st.sidebar.title("Navigation")
|
| 159 |
+
selection = st.sidebar.radio("Go to", ["Home", "Dataset Analysis", "Evaluation Results", "Run Inference", "Dissertation Report"])
|
| 160 |
|
| 161 |
if selection == "Home":
|
| 162 |
st.title("MultiModal Learning for Knowledg-Based Visual Question Answering")
|