Mirrors
/
EveryDream2trainer
mirror of https://github.com/victorchall/EveryDream2trainer.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

cleaner viewing for grounding ui

This commit is contained in:
Victor Hall 2024-03-03 16:55:30 -05:00
parent fdf230634e
commit a90e13a926
1 changed files with 11 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
scripts/grounding_ui.py
View File

@ -40,14 +40,13 @@ class ImageBoundingBoxApp:
img_path = os.path.join(self.folder_path, image_name)
ent_path = img_path.rsplit('.', 1)[0] + '.ent'
# Load image
image = cv2.imread(img_path)
image_cp = image.copy()
image_h, image_w, _ = image.shape
font_scale = max(0.5 * (image_h + image_w) / 1000,0.35)
thickness = max(int(0.5 * (image_h + image_w) / 500),1)
print(f"Font Scale: {font_scale}, Thickness: {thickness}")
shadow_thickness = 3
# Read .ent file
if os.path.exists(ent_path):
with open(ent_path, 'r') as f:
entities = ast.literal_eval(f.read())
@ -55,15 +54,17 @@ class ImageBoundingBoxApp:
for entity_name, _, bboxes in entities:
for (x1_norm, y1_norm, x2_norm, y2_norm) in bboxes:
orig_x1, orig_y1, orig_x2, orig_y2 = int(x1_norm * image_w), int(y1_norm * image_h), int(x2_norm * image_w), int(y2_norm * image_h)
color = (0, 255, 0) # Green color for bounding box
cv2.rectangle(image, (orig_x1, orig_y1), (orig_x2, orig_y2), color, 2)
#shadow text
cv2.putText(image, entity_name, (orig_x1, orig_y1 + int(30*font_scale)), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0,0,0), thickness+3)
#cv2.putText(image, entity_name, (orig_x1+1, orig_y1 + int(30*font_scale)-1), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0,0,0), thickness)
#text
rand_r = np.random.randint(0, 128)
rand_g = np.random.randint(144, 240)
rand_b = 128 - rand_r
color = (rand_r, rand_g, rand_b) # Green color for bounding box
cv2.rectangle(image, (orig_x1, orig_y1), (orig_x2, orig_y2), color, thickness)
cv2.putText(image, entity_name, (orig_x1, orig_y1 + int(30*font_scale)), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (0,0,0), thickness+shadow_thickness)
cv2.putText(image, entity_name, (orig_x1, orig_y1 + int(30*font_scale)), cv2.FONT_HERSHEY_SIMPLEX, font_scale, (255,255,255), thickness)
overlay_weight = 0.75
image = cv2.addWeighted(image, overlay_weight, image_cp, 1-overlay_weight, 0.0)
# Convert to PIL Image to display in Tkinter
image_pil = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
# Resize image to fit the canvas size while maintaining aspect ratio