Spaces:
Running
Running
add quad boxes
Browse files- app.py +32 -16
- requirements.txt +2 -1
app.py
CHANGED
@@ -1,6 +1,8 @@
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
-
|
|
|
|
|
4 |
from transformers import AutoProcessor
|
5 |
from modeling_florence2 import Florence2ForConditionalGeneration
|
6 |
import io
|
@@ -115,14 +117,15 @@ def fig_to_pil(fig):
|
|
115 |
def plot_bbox(image, data, use_quad_boxes=False):
|
116 |
fig, ax = plt.subplots()
|
117 |
ax.imshow(image)
|
118 |
-
|
119 |
# Handle both 'bboxes' and 'quad_boxes'
|
120 |
if use_quad_boxes:
|
121 |
for quad_box, label in zip(data.get('quad_boxes', []), data.get('labels', [])):
|
122 |
quad_box = np.array(quad_box).reshape(-1, 2)
|
123 |
-
poly =
|
124 |
ax.add_patch(poly)
|
125 |
-
plt.text(quad_box[0][0], quad_box[0][1], label, color='white', fontsize=8,
|
|
|
126 |
else:
|
127 |
bboxes = data.get('bboxes', [])
|
128 |
labels = data.get('labels', [])
|
@@ -149,49 +152,60 @@ def draw_ocr_bboxes(image, prediction):
|
|
149 |
fill=color)
|
150 |
return image
|
151 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
152 |
def process_image(image, task):
|
153 |
prompt = TASK_PROMPTS[task]
|
154 |
-
|
155 |
# Print the inputs for debugging
|
156 |
print(f"\n--- Processing Task: {task} ---")
|
157 |
print(f"Prompt: {prompt}")
|
158 |
-
|
159 |
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
|
160 |
-
|
161 |
# Print the input tensors for debugging
|
162 |
print(f"Model Input: {inputs}")
|
163 |
-
|
164 |
generated_ids = model.generate(
|
165 |
**inputs,
|
166 |
max_new_tokens=1024,
|
167 |
num_beams=3,
|
168 |
do_sample=False
|
169 |
)
|
170 |
-
|
171 |
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
|
172 |
-
|
173 |
# Print the raw generated output for debugging
|
174 |
print(f"Raw Model Output: {generated_text}")
|
175 |
-
|
176 |
parsed_answer = processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))
|
177 |
-
|
178 |
# Print the parsed answer for debugging
|
179 |
print(f"Parsed Answer: {parsed_answer}")
|
180 |
-
|
181 |
return parsed_answer
|
182 |
|
|
|
183 |
def main_process(image, task):
|
184 |
result = process_image(image, task)
|
185 |
|
186 |
if task in IMAGE_TASKS:
|
187 |
-
if task == "OCR with Region":
|
188 |
fig = plot_bbox(image, result.get('<OCR_WITH_REGION>', {}), use_quad_boxes=True)
|
189 |
output_image = fig_to_pil(fig)
|
190 |
text_output = result.get('<OCR_WITH_REGION>', {}).get('recognized_text', 'No text found')
|
191 |
-
|
192 |
# Debugging: Print the recognized text
|
193 |
print(f"Recognized Text: {text_output}")
|
194 |
-
|
195 |
return output_image, gr.update(visible=True), text_output, gr.update(visible=True)
|
196 |
else:
|
197 |
fig = plot_bbox(image, result.get(TASK_PROMPTS[task], {}))
|
@@ -219,6 +233,8 @@ with gr.Blocks(title="PLeIAs/📸📈✍🏻Florence-PDF") as iface:
|
|
219 |
output_image = gr.Image(label="PLeIAs/📸📈✍🏻Florence-PDF", visible=False)
|
220 |
output_text = gr.Textbox(label="PLeIAs/📸📈✍🏻Florence-PDF", visible=True)
|
221 |
|
|
|
|
|
222 |
def process_and_update(image, task):
|
223 |
if image is None:
|
224 |
return None, gr.update(visible=False), "Please upload an image first.", gr.update(visible=True)
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
+
import cv2
|
4 |
+
import numpy as np
|
5 |
+
from matplotlib import pyplot as pltfrom PIL import Image, ImageDraw
|
6 |
from transformers import AutoProcessor
|
7 |
from modeling_florence2 import Florence2ForConditionalGeneration
|
8 |
import io
|
|
|
117 |
def plot_bbox(image, data, use_quad_boxes=False):
|
118 |
fig, ax = plt.subplots()
|
119 |
ax.imshow(image)
|
120 |
+
|
121 |
# Handle both 'bboxes' and 'quad_boxes'
|
122 |
if use_quad_boxes:
|
123 |
for quad_box, label in zip(data.get('quad_boxes', []), data.get('labels', [])):
|
124 |
quad_box = np.array(quad_box).reshape(-1, 2)
|
125 |
+
poly = Polygon(quad_box, linewidth=1, edgecolor='r', facecolor='none')
|
126 |
ax.add_patch(poly)
|
127 |
+
plt.text(quad_box[0][0], quad_box[0][1], label, color='white', fontsize=8,
|
128 |
+
bbox=dict(facecolor='red', alpha=0.5))
|
129 |
else:
|
130 |
bboxes = data.get('bboxes', [])
|
131 |
labels = data.get('labels', [])
|
|
|
152 |
fill=color)
|
153 |
return image
|
154 |
|
155 |
+
def draw_bounding_boxes(image, quad_boxes, labels, color=(0, 255, 0), thickness=2):
|
156 |
+
"""
|
157 |
+
Draws quadrilateral bounding boxes on the image.
|
158 |
+
|
159 |
+
Args:
|
160 |
+
image: The original image where the bounding boxes will be drawn.
|
161 |
+
quad_boxes: List of quadrilateral bounding box points. Each bounding box contains four points.
|
162 |
+
labels: List of labels corresponding to each bounding box.
|
163 |
+
color: Color of the bounding box. Default is green.
|
164 |
+
thickness: Thickness of the bounding box lines. Default is 2.
|
165 |
+
"""
|
166 |
+
for i, quad in enumerate(quad_boxes):
|
167 |
+
points = np.array(quad, dtype=np.int32).reshape((-1, 1, 2)) # Reshape the quad points for drawing
|
168 |
+
image = cv2.polylines(image, [points], isClosed=True, color=color, thickness=thickness)
|
169 |
+
# Add label text near the top-left point of the bounding box
|
170 |
+
label_pos = (int(quad[0]), int(quad[1]) - 10) # Positioning label slightly above the bounding box
|
171 |
+
cv2.putText(image, labels[i], label_pos, cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, thickness)
|
172 |
+
|
173 |
+
return image
|
174 |
+
|
175 |
def process_image(image, task):
|
176 |
prompt = TASK_PROMPTS[task]
|
|
|
177 |
# Print the inputs for debugging
|
178 |
print(f"\n--- Processing Task: {task} ---")
|
179 |
print(f"Prompt: {prompt}")
|
|
|
180 |
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
|
|
|
181 |
# Print the input tensors for debugging
|
182 |
print(f"Model Input: {inputs}")
|
|
|
183 |
generated_ids = model.generate(
|
184 |
**inputs,
|
185 |
max_new_tokens=1024,
|
186 |
num_beams=3,
|
187 |
do_sample=False
|
188 |
)
|
189 |
+
|
190 |
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
|
|
|
191 |
# Print the raw generated output for debugging
|
192 |
print(f"Raw Model Output: {generated_text}")
|
|
|
193 |
parsed_answer = processor.post_process_generation(generated_text, task=prompt, image_size=(image.width, image.height))
|
|
|
194 |
# Print the parsed answer for debugging
|
195 |
print(f"Parsed Answer: {parsed_answer}")
|
|
|
196 |
return parsed_answer
|
197 |
|
198 |
+
|
199 |
def main_process(image, task):
|
200 |
result = process_image(image, task)
|
201 |
|
202 |
if task in IMAGE_TASKS:
|
203 |
+
if task == "📸✍🏻OCR with Region":
|
204 |
fig = plot_bbox(image, result.get('<OCR_WITH_REGION>', {}), use_quad_boxes=True)
|
205 |
output_image = fig_to_pil(fig)
|
206 |
text_output = result.get('<OCR_WITH_REGION>', {}).get('recognized_text', 'No text found')
|
|
|
207 |
# Debugging: Print the recognized text
|
208 |
print(f"Recognized Text: {text_output}")
|
|
|
209 |
return output_image, gr.update(visible=True), text_output, gr.update(visible=True)
|
210 |
else:
|
211 |
fig = plot_bbox(image, result.get(TASK_PROMPTS[task], {}))
|
|
|
233 |
output_image = gr.Image(label="PLeIAs/📸📈✍🏻Florence-PDF", visible=False)
|
234 |
output_text = gr.Textbox(label="PLeIAs/📸📈✍🏻Florence-PDF", visible=True)
|
235 |
|
236 |
+
gr.Markdown(model_presentation)
|
237 |
+
|
238 |
def process_and_update(image, task):
|
239 |
if image is None:
|
240 |
return None, gr.update(visible=False), "Please upload an image first.", gr.update(visible=True)
|
requirements.txt
CHANGED
@@ -3,4 +3,5 @@ transformers
|
|
3 |
accelerate
|
4 |
pillow
|
5 |
einops
|
6 |
-
timm
|
|
|
|
3 |
accelerate
|
4 |
pillow
|
5 |
einops
|
6 |
+
timm
|
7 |
+
opencv-python
|