import copy import os import sys sys.path.append('src') from collections import defaultdict from functools import lru_cache import gradio as gr import matplotlib.pyplot as plt import numpy as np import pandas as pd import torch from deep_translator import GoogleTranslator from Nets import CustomResNet18 from PIL import Image from torchcam.methods import GradCAM, GradCAMpp, SmoothGradCAMpp, XGradCAM from torchcam.utils import overlay_mask from torchvision.transforms.functional import to_pil_image from tqdm import tqdm from util import transform from gradio_blocks import build_video_to_camvideo import cv2 import ffmpeg import shutil import mediapy ffmpeg_path = shutil.which('ffmpeg') mediapy.set_ffmpeg(ffmpeg_path) IMAGE_PATH = os.path.join(os.getcwd(), 'src/examples') IMAGES_PER_ROW = 10 MAXIMAL_FRAMES = 1000 BATCHES_TO_PROCESS = 15 OUTPUT_FPS = 10 MAX_OUT_FRAMES = 70 CAM_METHODS = { "GradCAM": GradCAM, "GradCAM++": GradCAMpp, "XGradCAM": XGradCAM, "SmoothGradCAM++": SmoothGradCAMpp, } model = CustomResNet18(90).eval() model.load_state_dict(torch.load('src/results/models/best_model.pth', map_location=torch.device('cpu'))) cam_model = copy.deepcopy(model) data_df = pd.read_csv('src/cache/val_df.csv') C_NUM_TO_NAME = data_df[['encoded_target', 'target']].drop_duplicates().sort_values('encoded_target').set_index('encoded_target')['target'].to_dict() C_NAME_TO_NUM = {v: k for k, v in C_NUM_TO_NAME.items()} ALL_CLASSES = sorted(list(C_NUM_TO_NAME.values()), key=lambda x: x.lower()) def get_class_name(idx): return C_NUM_TO_NAME[idx] def get_class_idx(name): return C_NAME_TO_NUM[name] @lru_cache(maxsize=100) def get_translated(to_translate): # return "ssss" return GoogleTranslator(source="en", target="de").translate(to_translate) for idx in range(90): get_translated(get_class_name(idx)) def infer_image(image, image_sketch): image = image if image is not None else image_sketch image = transform(image) image = image.unsqueeze(0) with torch.no_grad(): output = model(image) distribution = torch.nn.functional.softmax(output, dim=1) ret = defaultdict(float) for idx, prob in enumerate(distribution[0]): animal = f'{get_class_name(idx)} ({get_translated(get_class_name(idx))})' ret[animal] = prob.item() return ret def gradcam(image, image_sketch=None, alpha=0.5, cam_method=GradCAM, layer=None, specific_class="Predicted Class"): image = image if image is not None else image_sketch if layer == 'layer1': layers = [model.resnet.layer1] elif layer == 'layer2': layers = [model.resnet.layer2] elif layer == 'layer3': layers = [model.resnet.layer3] elif layer == 'layer4': layers = [model.resnet.layer4] else: layers = [model.resnet.layer1, model.resnet.layer2, model.resnet.layer3, model.resnet.layer4] model.eval() img_tensor = transform(image).unsqueeze(0) cam = CAM_METHODS[cam_method](model, target_layer=layers) output = model(img_tensor) class_to_explain = output.squeeze(0).argmax().item() if specific_class == "Predicted Class" else get_class_idx(specific_class) activation_map = cam(class_to_explain, output) result = overlay_mask(image, to_pil_image(activation_map[0].squeeze(0), mode='F'), alpha=alpha) cam.remove_hooks() # # height maximal 300px # if result.size[1] > 300: # ratio = 300 / result.size[1] # result = result.resize((int(result.size[0] * ratio), 300)) return result def gradcam_video(video, alpha=0.5, cam_method=GradCAM, layer=None, specific_class="Predicted Class"): global OUTPUT_FPS, MAXIMAL_FRAMES, BATCHES_TO_PROCESS, MAX_OUT_FRAMES video = cv2.VideoCapture(video) fps = int(video.get(cv2.CAP_PROP_FPS)) if OUTPUT_FPS == -1: OUTPUT_FPS = fps width = int(video.get(cv2.CAP_PROP_FRAME_WIDTH)) height = int(video.get(cv2.CAP_PROP_FRAME_HEIGHT)) if width > 3000 or height > 3000: raise gr.Error("The video is too big. The maximal size is 3000x3000.") print(f'FPS: {fps}, Width: {width}, Height: {height}') frames = list() success, image = video.read() while success: frames.append(image) success, image = video.read() print(f'Frames: {len(frames)}') if len(frames) == 0: raise gr.Error("The video is empty.") if len(frames) >= MAXIMAL_FRAMES: raise gr.Error(f"The video is too long. The maximal length is {MAXIMAL_FRAMES} frames.") if len(frames) > MAX_OUT_FRAMES: frames = frames[::len(frames) // MAX_OUT_FRAMES] print(f'Frames to process: {len(frames)}') processed = [Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)) for frame in frames] # generate lists in lists for the images for batch processing. 10 images per inner list.. batched = [processed[i:i + BATCHES_TO_PROCESS] for i in range(0, len(processed), BATCHES_TO_PROCESS)] model.eval() if layer == 'layer1': layers = [model.resnet.layer1] elif layer == 'layer2': layers = [model.resnet.layer2] elif layer == 'layer3': layers = [model.resnet.layer3] elif layer == 'layer4': layers = [model.resnet.layer4] else: layers = [model.resnet.layer1, model.resnet.layer2, model.resnet.layer3, model.resnet.layer4] cam = CAM_METHODS[cam_method](model, target_layer=layers) results = list() for i, batch in enumerate(tqdm(batched)): images_tensor = torch.stack([transform(image) for image in batch]) outputs = model(images_tensor) out_classes = [output.argmax().item() for output in outputs] classes_to_explain = out_classes if specific_class == "Predicted Class" else [get_class_idx(specific_class)] * len(out_classes) activation_maps = cam(classes_to_explain, outputs) for j, activation_map in enumerate(activation_maps[0]): result = overlay_mask(batch[j], to_pil_image(activation_map, mode='F'), alpha=alpha) results.append(cv2.cvtColor(np.array(result), cv2.COLOR_RGB2BGR)) cam.remove_hooks() # save video # fourcc = cv2.VideoWriter_fourcc(*'AVC1') # fourcc = cv2.VideoWriter_fourcc(*'MP4V') # fourcc = cv2.VideoWriter_fourcc(*'XVID') # size = (results[0].shape[1], results[0].shape[0]) # video = cv2.VideoWriter('src/results/gradcam_video.mp4', fourcc, OUTPUT_FPS, size) # for frame in results: # video.write(frame) mediapy.write_video('src/results/gradcam_video.mp4', results, fps=OUTPUT_FPS) video.release() return 'src/results/gradcam_video.mp4' def load_examples(): folder_name_to_header = { "AI_Generated": "AI Generated Images", "true_predicted": "True Predicted Images (Validation Set)", "false_predicted": "False Predicted Images (Validation Set)", "others": "Other interesting images from the internet" } images_description = { "AI_Generated": "These images are generated by Dalle3 and Stable Diffusion. All of them are not real images and because of that it is interesting to see how the model predicts them.", "true_predicted": "These images are from the validation set and the model predicted them correctly.", "false_predicted": "These images are from the validation set and the model predicted them incorrectly. Maybe you can see why the model predicted them incorrectly using the GradCAM visualization. :)", "others": "These images are from the internet and are not part of the validation set. They are interesting because most of them show different animals." } loaded_images = defaultdict(list) for image_type in ["AI_Generated", "true_predicted", "false_predicted", "others"]: # for image_type in os.listdir(IMAGE_PATH): full_path = os.path.join(IMAGE_PATH, image_type).replace('\\', '/').replace('//', '/') gr.Markdown(f'## {folder_name_to_header[image_type]}') gr.Markdown(images_description[image_type]) images_to_load = os.listdir(full_path) rows = (len(images_to_load) // IMAGES_PER_ROW) + 1 for i in range(rows): with gr.Row(elem_classes=["row-example-images"], equal_height=False): for j in range(IMAGES_PER_ROW): if i * IMAGES_PER_ROW + j >= len(images_to_load): break image = images_to_load[i * IMAGES_PER_ROW + j] loaded_images[image_type].append( gr.Image( value=os.path.join(full_path, image), label=f"image ({get_translated(image.split('.')[0])})", type="pil", interactive=False, elem_classes=["selectable_images"], ) ) return loaded_images css = """ #logo {text-align: right;} p {text-align: justify; text-justify: inter-word; font-size: 1.1em; line-height: 1.2em;} .svelte-1btp92j.selectable {cursor: pointer !important; } """ with gr.Blocks(theme='freddyaboulton/dracula_revamped', css=css) as demo: # ------------------------------------------- # HEADER WITH LOGO # ------------------------------------------- with gr.Row(): with open('src/header.md', 'r', encoding='utf-8') as f: markdown_string = f.read() with gr.Column(scale=10): header = gr.Markdown(markdown_string) with gr.Column(scale=1): pil_logo = Image.open('animals.png') logo = gr.Image(value=pil_logo, scale=2, interactive=False, show_download_button=False, show_label=False, container=False, elem_id="logo") # ------------------------------------------- # INPUT IMAGE # ------------------------------------------- with gr.Row(): with gr.Tab("Upload Image"): with gr.Row(variant="panel", equal_height=True): user_image = gr.Image( type="pil", label="Upload Your Own Image", info="You can also upload your own image for prediction.", ) with gr.Tab("Draw Image"): with gr.Row(variant="panel", equal_height=True): user_image_sketched = gr.Image( type="pil", source="canvas", tool="color-sketch", label="Draw Your Own Image", info="You can also draw your own image for prediction.", ) # ------------------------------------------- # TOOLS # ------------------------------------------- with gr.Row(): # ------------------------------------------- # PREDICT # ------------------------------------------- with gr.Tab("Predict"): with gr.Column(): output = gr.Label( num_top_classes=5, label="Output", info="Top three predicted classes and their confidences.", scale=5, ) predict_mode_button = gr.Button(value="Predict Animal", label="Predict", info="Click to make a prediction.", scale=1) predict_mode_button.click(fn=infer_image, inputs=[user_image, user_image_sketched], outputs=output, queue=True) # ------------------------------------------- # EXPLAIN # ------------------------------------------- with gr.Tab("Explain Image"): with gr.Row(): with gr.Column(): cam_method = gr.Radio( list(CAM_METHODS.keys()), label="GradCAM Method", value="GradCAM", interactive=True, scale=2, ) cam_method.description = "Here you can choose the GradCAM method." cam_method.description_place = "left" alpha = gr.Slider( minimum=.1, maximum=.9, value=0.5, interactive=True, step=.1, label="Alpha", scale=1, ) alpha.description = "Here you can choose the alpha value." alpha.description_place = "left" layer = gr.Radio( ["layer1", "layer2", "layer3", "layer4", "all"], label="Layer", value="layer4", interactive=True, scale=2, ) layer.description = "Here you can choose the layer to visualize." layer.description_place = "left" animal_to_explain = gr.Dropdown( choices=["Predicted Class"] + ALL_CLASSES, label="Animal", value="Predicted Class", interactive=True, scale=2, ) animal_to_explain.description = "Here you can choose the animal to explain. If you choose 'Predicted Class' the method will explain the predicted class." animal_to_explain.description_place = "center" with gr.Column(): output_cam = gr.Image( type="pil", label="GradCAM", info="GradCAM visualization" ) gradcam_mode_button = gr.Button(value="Show GradCAM", label="GradCAM", info="Click to make a prediction.", scale=1) gradcam_mode_button.click(fn=gradcam, inputs=[user_image, user_image_sketched, alpha, cam_method, layer, animal_to_explain], outputs=output_cam, queue=True) # ------------------------------------------- # Video CAM # ------------------------------------------- with gr.Tab("Explain Video"): build_video_to_camvideo(CAM_METHODS, ALL_CLASSES, gradcam_video) # ------------------------------------------- # EXAMPLES # ------------------------------------------- with gr.Tab("Example Images"): placeholder = gr.Markdown("## Example Images") loaded_images = load_examples() for k in loaded_images.keys(): for image in loaded_images[k]: image.select(fn=lambda x: x, inputs=[image], outputs=[user_image]) if __name__ == "__main__": demo.queue() demo.launch()