Update app.py

app.py

@@ -5,7 +5,7 @@ import gradio as gr
 import os
 import xml.etree.ElementTree as ET
 
-# ---------------- Helper Functions ----------------
+# ---------------- Helper functions ----------------
 def get_rotated_rect_corners(x, y, w, h, rotation_deg):
     rot_rad = np.deg2rad(rotation_deg)
     cos_r, sin_r = np.cos(rot_rad), np.sin(rot_rad)
@@ -45,6 +45,18 @@ def parse_xml_points(xml_file):
         points.append([float(elem.get("x")), float(elem.get("y"))])
     return np.array(points,dtype=np.float32).reshape(-1,2)
 
+# ---------------- Padding Helper ----------------
+def pad_to_size(img, target_h, target_w):
+    h, w = img.shape[:2]
+    top_pad = 0
+    left_pad = 0
+    bottom_pad = target_h - h
+    right_pad = target_w - w
+    # Create white canvas
+    canvas = np.ones((target_h, target_w,3), dtype=np.uint8)*255
+    canvas[top_pad:top_pad+h, left_pad:left_pad+w] = img
+    return canvas
+
 # ---------------- Main Function ----------------
 def homography_all_detectors(flat_file, persp_file, json_file, xml_file):
     flat_img = cv2.imread(flat_file)
@@ -60,17 +72,15 @@ def homography_all_detectors(flat_file, persp_file, json_file, xml_file):
     xml_points = parse_xml_points(xml_file.name)
 
     methods = ["SIFT","ORB","BRISK","KAZE","AKAZE"]
-    gallery_images = []
+    gallery_paths = []
     download_files = []
 
     for method in methods:
         kp1,kp2,good_matches = detect_and_match(flat_gray,persp_gray,method)
         if kp1 is None or kp2 is None or len(good_matches)<4: continue
 
-        # Feature matching
         match_img = cv2.drawMatches(flat_img,kp1,persp_img,kp2,good_matches,None,flags=2)
 
-        # Homography & ROI
         src_pts = np.float32([kp1[m.queryIdx].pt for m in good_matches]).reshape(-1,1,2)
         dst_pts = np.float32([kp2[m.trainIdx].pt for m in good_matches]).reshape(-1,1,2)
         H,_ = cv2.findHomography(src_pts,dst_pts,cv2.RANSAC,5.0)
@@ -86,38 +96,36 @@ def homography_all_detectors(flat_file, persp_file, json_file, xml_file):
         xml_mapped = cv2.perspectiveTransform(xml_points.reshape(-1,1,2),H).reshape(-1,2)
         for px,py in xml_mapped: cv2.circle(xml_gt_img,(int(px),int(py)),5,(0,0,255),-1)
 
-        # --------- For display only: resize to match grid ---------
-        def resize_to_height(img, target_h):
-            h, w = img.shape[:2]
-            scale = target_h / h
-            new_w = int(w * scale)
-            return cv2.resize(img, (new_w, target_h))
-
-        target_h = 300  # temporary display height for gallery
-        flat_disp = resize_to_height(flat_img, target_h)
-        match_disp = resize_to_height(match_img, target_h)
-        roi_disp = resize_to_height(persp_roi, target_h)
-        xml_disp = resize_to_height(xml_gt_img, target_h)
-
-        # Merge 2x2 grid for gallery display
-        top = np.hstack([flat_disp, match_disp])
-        bottom = np.hstack([roi_disp, xml_disp])
+        # Convert to RGB
+        flat_rgb = cv2.cvtColor(flat_img,cv2.COLOR_BGR2RGB)
+        match_rgb = cv2.cvtColor(match_img,cv2.COLOR_BGR2RGB)
+        roi_rgb = cv2.cvtColor(persp_roi,cv2.COLOR_BGR2RGB)
+        xml_rgb = cv2.cvtColor(xml_gt_img,cv2.COLOR_BGR2RGB)
+
+        # Determine max height and width for padding
+        max_h = max(flat_rgb.shape[0], match_rgb.shape[0], roi_rgb.shape[0], xml_rgb.shape[0])
+        max_w = max(flat_rgb.shape[1], match_rgb.shape[1], roi_rgb.shape[1], xml_rgb.shape[1])
+
+        # Pad all images to same size
+        flat_pad = pad_to_size(flat_rgb, max_h, max_w)
+        match_pad = pad_to_size(match_rgb, max_h, max_w)
+        roi_pad = pad_to_size(roi_rgb, max_h, max_w)
+        xml_pad = pad_to_size(xml_rgb, max_h, max_w)
+
+        # Merge 2x2 grid
+        top = np.hstack([flat_pad, match_pad])
+        bottom = np.hstack([roi_pad, xml_pad])
         combined_grid = np.vstack([top, bottom])
-        gallery_images.append((combined_grid,f"{method} Detector"))
 
-        # Save original resolution for download
+        # Save combined grid
         base_name = os.path.splitext(os.path.basename(persp_file))[0]
         file_name = f"{base_name}_{method.lower()}.png"
-        # Merge original images for download
-        top_orig = np.hstack([flat_img, match_img])
-        bottom_orig = np.hstack([persp_roi, xml_gt_img])
-        combined_orig = np.vstack([top_orig, bottom_orig])
-        cv2.imwrite(file_name, combined_orig)
+        cv2.imwrite(file_name, cv2.cvtColor(combined_grid,cv2.COLOR_RGB2BGR))
+        gallery_paths.append(file_name)
         download_files.append(file_name)
 
-    # Ensure 5 outputs
     while len(download_files)<5: download_files.append(None)
-    return gallery_images, download_files[0], download_files[1], download_files[2], download_files[3], download_files[4]
+    return gallery_paths, download_files[0], download_files[1], download_files[2], download_files[3], download_files[4]
 
 # ---------------- Gradio UI ----------------
 iface = gr.Interface(
@@ -137,7 +145,7 @@ iface = gr.Interface(
         gr.File(label="Download AKAZE Result")
     ],
     title="Homography ROI Projection with Feature Matching & XML GT",
-    description="Shows 4 views per detector. Grid resized for display, original resolution saved for download."
+    description="Flat + Perspective images with mockup.json & XML. Original resolution maintained. Grid aligned with white padding."
 )
 
 iface.launch()