Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| import cv2 | |
| import numpy as np | |
| from PIL import Image | |
| import io | |
| st.set_page_config(layout="wide") | |
| st.title("Image Augmentation Tool") | |
| # Set max file size limit in bytes (e.g., 5MB) | |
| MAX_FILE_SIZE = 5 * 1024 * 1024 # 5MB | |
| uploaded_files = st.file_uploader( | |
| "Upload one or more images (Max 5MB each)", | |
| type=["jpg", "jpeg", "png"], | |
| accept_multiple_files=True | |
| ) | |
| # Function to apply all transformations | |
| def apply_all_transformations(img_array): | |
| results = [] | |
| gray = cv2.cvtColor(img_array, cv2.COLOR_RGB2GRAY) | |
| results.append(("Grayscale", gray)) | |
| rotated = cv2.rotate(img_array, cv2.ROTATE_90_CLOCKWISE) | |
| results.append(("Rotate 90°", rotated)) | |
| flip_h = cv2.flip(img_array, 1) | |
| results.append(("Flip Horizontally", flip_h)) | |
| flip_v = cv2.flip(img_array, 0) | |
| results.append(("Flip Vertically", flip_v)) | |
| rows, cols = img_array.shape[:2] | |
| shear_M = np.float32([[1, 0.5, 0], [0.2, 1, 0]]) | |
| sheared = cv2.warpAffine(img_array, shear_M, (cols, rows)) | |
| results.append(("Shearing", sheared)) | |
| trans_M = np.float32([[1, 0, 50], [0, 1, 30]]) | |
| translated = cv2.warpAffine(img_array, trans_M, (cols, rows)) | |
| results.append(("Translation", translated)) | |
| h, w = img_array.shape[:2] | |
| x_start = w // 4 | |
| x_end = 3 * x_start | |
| y_start = h // 4 | |
| y_end = 3 * y_start | |
| cropped = img_array[y_start:y_end, x_start:x_end] | |
| results.append(("Crop (center 50%)", cropped, (x_start, x_end, y_start, y_end))) | |
| return results | |
| if uploaded_files: | |
| for uploaded_file in uploaded_files: | |
| # Check file size limit | |
| if uploaded_file.size > MAX_FILE_SIZE: | |
| st.error(f"❌ File {uploaded_file.name} is too large. Max size allowed is 5MB.") | |
| continue | |
| st.markdown("---") | |
| st.subheader(f"📷 Original: {uploaded_file.name}") | |
| image = Image.open(uploaded_file).convert("RGB") | |
| img_array = np.array(image) | |
| # Resize original image if width exceeds 400px | |
| max_display_width = 400 | |
| orig_w, orig_h = image.size | |
| if orig_w > max_display_width: | |
| new_h = int(orig_h * max_display_width / orig_w) | |
| image_resized = image.resize((max_display_width, new_h)) | |
| else: | |
| image_resized = image | |
| # Show resized original image | |
| st.image(image_resized, caption="Original Image", use_container_width=False) | |
| # Apply all transformations | |
| transformations = apply_all_transformations(img_array) | |
| st.markdown("### 🧪 Transformed Versions") | |
| # Create columns dynamically (3 per row) | |
| cols = st.columns(3) | |
| for idx, item in enumerate(transformations): | |
| name = item[0] | |
| transformed = item[1] | |
| coords = item[2] if len(item) > 2 else None | |
| col = cols[idx % 3] | |
| with col: | |
| st.markdown(f"**{name}**") | |
| if len(transformed.shape) == 2: | |
| st.image(transformed, caption=name, use_container_width=True, channels="GRAY") | |
| save_img = cv2.cvtColor(transformed, cv2.COLOR_GRAY2RGB) | |
| else: | |
| st.image(transformed, caption=name, use_container_width=True) | |
| save_img = transformed | |
| # Save transformed image to bytes | |
| result_img = Image.fromarray(save_img) | |
| buf = io.BytesIO() | |
| result_img.save(buf, format="PNG") | |
| byte_data = buf.getvalue() | |
| st.download_button( | |
| label="⬇️ Download", | |
| data=byte_data, | |
| file_name=f"{uploaded_file.name.split('.')[0]}_{name.replace(' ', '_').lower()}.png", | |
| mime="image/png" | |
| ) | |