initial2

Interface_Dependencies/plaus_functs.py

@@ -0,0 +1,124 @@
+## Current Implementation of Smooth_grad (Allows more choices)
+
+import torch
+import numpy as np
+from plot_functs import *    
+
+def generate_vanilla_grad(model, input_tensor, loss_func = None, 
+                          targets=None, metric=None, out_num = 1, 
+                          norm=False, device='cpu'):    
+    """
+    Computes the vanilla gradient of the input tensor with respect to the output of the given model.
+
+    Args:
+        model (torch.nn.Module): The model to compute the gradient with respect to.
+        input_tensor (torch.Tensor): The input tensor to compute the gradient for.
+        loss_func (callable, optional): The loss function to use. If None, the gradient is computed with respect to the output tensor.
+        targets (torch.Tensor, optional): The target tensor to use with the loss function. Defaults to None.
+        metric (callable, optional): The metric function to use with the loss function. Defaults to None.
+        out_num (int, optional): The index of the output tensor to compute the gradient with respect to. Defaults to 1.
+        norm (bool, optional): Whether to normalize the attribution map. Defaults to False.
+        device (str, optional): The device to use for computation. Defaults to 'cpu'.
+
+    Returns:
+        torch.Tensor: The attribution map computed as the gradient of the input tensor with respect to the output tensor.
+    """
+    # maybe add model.train() at the beginning and model.eval() at the end of this function
+
+    # Set requires_grad attribute of tensor. Important for computing gradients
+    input_tensor.requires_grad = True
+    
+    # Zero gradients
+    model.zero_grad()
+
+    # Forward pass
+    train_out = model(input_tensor) # training outputs (no inference outputs in train mode)
+    
+    # train_out[1] = torch.Size([4, 3, 80, 80, 7]) HxWx(#anchorxC) cls (class probabilities)
+    # train_out[0] = torch.Size([4, 3, 160, 160, 7]) HxWx(#anchorx4) reg (location and scaling)
+    # train_out[2] = torch.Size([4, 3, 40, 40, 7]) HxWx(#anchorx1) obj (objectness score or confidence)
+    
+    out_num = out_num - 1
+    
+    if loss_func is None:
+        grad_wrt = train_out[out_num]
+        grad_wrt_outputs = torch.ones_like(grad_wrt)
+    else:
+        loss, loss_items = loss_func(train_out, targets.to(device), input_tensor, metric=metric)  # loss scaled by batch_size
+        grad_wrt = loss
+        grad_wrt_outputs = None
+        # loss.backward(retain_graph=True, create_graph=True)
+        # gradients = input_tensor.grad
+    
+    gradients = torch.autograd.grad(grad_wrt, input_tensor, 
+                                        grad_outputs=grad_wrt_outputs, 
+                                        retain_graph=True, create_graph=True)
+
+    # Convert gradients to numpy array
+    gradients = gradients[0].detach().cpu().numpy()
+
+    if norm:
+        # Take absolute values of gradients
+        gradients = np.absolute(gradients)
+
+        # Sum across color channels
+        attribution_map = np.sum(gradients, axis=0)
+
+        # Normalize attribution map
+        attribution_map /= np.max(attribution_map)
+    else:
+        # Sum across color channels
+        attribution_map = gradients
+
+    # Set model back to training mode
+    # model.train()
+    
+    return torch.tensor(attribution_map, dtype=torch.float32, device=device)
+
+
+def eval_plausibility(imgs, targets, attr_tensor, device):
+    """
+    Evaluate the plausibility of an object detection prediction by computing the Intersection over Union (IoU) between
+    the predicted bounding box and the ground truth bounding box.
+
+    Args:
+        im0 (numpy.ndarray): The input image.
+        targets (list): A list of targets, where each target is a list containing the class label and the ground truth
+            bounding box coordinates in the format [class_label, x1, y1, x2, y2].
+        attr (torch.Tensor): A tensor containing the normalized attribute values for the predicted
+            bounding box.
+
+    Returns:
+        float: The total IoU score for all predicted bounding boxes.
+    """
+    # if len(targets) == 0:
+    #     return 0
+    # MIGHT NEED TO NORMALIZE OR TAKE ABS VAL OF ATTR
+    # ALSO MIGHT NORMALIZE FOR THE SIZE OF THE BBOX
+    eval_totals = 0
+    eval_individual_data = []
+    targets_ = [[targets[i] for i in range(len(targets)) if int(targets[i][0]) == j] for j in range(int(max(targets[:,0])))]
+    for i, im0 in enumerate(imgs):
+        if len(targets[i]) == 0:
+            eval_individual_data.append([torch.tensor(0).to(device),])
+        else:
+            IoU_list = []
+            xyxy_pred = targets[i][2:] # * torch.tensor([im0.shape[2], im0.shape[1], im0.shape[2], im0.shape[1]])
+            xyxy_center = corners_coords(xyxy_pred) * torch.tensor([im0.shape[1], im0.shape[2], im0.shape[1], im0.shape[2]])
+            c1, c2 = (int(xyxy_center[0]), int(xyxy_center[1])), (int(xyxy_center[2]), int(xyxy_center[3]))
+            attr = normalize_tensor(abs(attr_tensor[i].clone().detach()))
+            IoU_num = (torch.sum(attr[:,c1[1]:c2[1], c1[0]:c2[0]]))
+            IoU_denom = (torch.sum(attr))
+            IoU = IoU_num / IoU_denom
+            IoU_list.append(IoU)
+        eval_totals += torch.mean(torch.tensor(IoU_list))
+        eval_individual_data.append(IoU_list)
+
+    return torch.tensor(eval_totals).requires_grad_(True)
+
+def corners_coords(center_xywh):
+    center_x, center_y, w, h = center_xywh
+    x = center_x - w/2
+    y = center_y - h/2
+    return torch.tensor([x, y, x+w, y+h])
+    

Interface_Dependencies/plot_functs.py

@@ -0,0 +1,97 @@
+import numpy as np
+import matplotlib.pyplot as plt
+import torch
+import torch.nn as nn
+
+
+def VisualizeNumpyImageGrayscale(image_3d):
+    r"""Returns a 3D tensor as a grayscale normalized between 0 and 1 2D tensor.
+    """
+    vmin = np.min(image_3d)
+    image_2d = image_3d - vmin
+    vmax = np.max(image_2d)
+    return (image_2d / vmax)
+
+def normalize_tensor(image_3d): 
+    r"""Returns a 3D tensor as a grayscale normalized between 0 and 1 2D tensor.
+    """
+    vmin = torch.min(image_3d)
+    image_2d = image_3d - vmin
+    vmax = torch.max(image_2d)
+    return (image_2d / vmax)
+
+def format_img(img_):
+    img_ = img_     # unnormalize
+    np_img = img_.numpy()
+    tp_img = np.transpose(np_img, (1, 2, 0))
+    return tp_img
+
+def imshow(img, save_path=None):
+    img = img     # unnormalize
+    try:
+        npimg = img.cpu().detach().numpy()
+    except:
+        npimg = img
+    tpimg = np.transpose(npimg, (1, 2, 0))
+    plt.imshow(tpimg)
+    if save_path != None:
+        plt.savefig(str(str(save_path) + ".png"))
+    #plt.show()
+
+def imshow_img(img, imsave_path):
+    # works for tensors and numpy arrays
+    try:
+        npimg = VisualizeNumpyImageGrayscale(img.numpy())
+    except:
+        npimg = VisualizeNumpyImageGrayscale(img)
+    npimg = np.transpose(npimg, (2, 0, 1))
+    imshow(npimg, save_path=imsave_path)
+    print("Saving image as ", imsave_path)
+    
+def returnGrad(img, labels, model, compute_loss, loss_metric, augment=None, device = 'cpu'):
+    model.train()
+    model.to(device)
+    img = img.to(device)
+    img.requires_grad_(True)
+    labels.to(device).requires_grad_(True)
+    model.requires_grad_(True)
+    cuda = device.type != 'cpu'
+    scaler = amp.GradScaler(enabled=cuda)
+    pred = model(img)
+    # out, train_out = model(img, augment=augment)  # inference and training outputs
+    loss, loss_items = compute_loss(pred, labels, metric=loss_metric)#[1][:3]  # box, obj, cls
+    # loss = criterion(pred, torch.tensor([int(torch.max(pred[0], 0)[1])]).to(device))
+    # loss = torch.sum(loss).requires_grad_(True)
+    
+    with torch.autograd.set_detect_anomaly(True):
+        scaler.scale(loss).backward(inputs=img)
+    # loss.backward()
+    
+#    S_c = torch.max(pred[0].data, 0)[0]
+    Sc_dx = img.grad
+    model.eval()
+    Sc_dx = torch.tensor(Sc_dx, dtype=torch.float32)
+    return Sc_dx
+
+def calculate_snr(img, attr, dB=True):
+    try:
+        img_np = img.detach().cpu().numpy()
+        attr_np = attr.detach().cpu().numpy()
+    except:
+        img_np = img
+        attr_np = attr
+    
+    # Calculate the signal power
+    signal_power = np.mean(img_np**2)
+
+    # Calculate the noise power
+    noise_power = np.mean(attr_np**2)
+
+    if dB == True:
+        # Calculate SNR in dB
+        snr = 10 * np.log10(signal_power / noise_power)
+    else:
+        # Calculate SNR
+        snr = signal_power / noise_power
+
+    return snr

Interface_Dependencies/run_methods.py

@@ -0,0 +1,156 @@
+import torch
+import os
+from PIL import Image
+import argparse
+
+import sys
+sys.path.append('Interface_Dependencies')
+sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface/Interface_Dependencies')
+sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface/yolov7-main')
+sys.path.append('./')  # to run '$ python *.py' files in subdirectories
+
+from ourDetect import detect, generate_feature_maps # used for output generation
+from utils.general import strip_optimizer # used for opt creation
+
+
+def correct_video(video):
+    """
+    Takes a video file of any type and turns it into a gradio compatible .mp4/264 video
+
+    Args:
+        video (str): The file path of the input video
+
+    Returns:
+        str: The file path of the output video
+    """
+    os.system("ffmpeg -i {file_str} -y -vcodec libx264 -acodec aac {file_str}.mp4".format(file_str = video))
+    return video+".mp4"
+
+def run_all(source_type, im, vid, src, inf_size=640, obj_conf_thr=0.25, iou_thr=0.45, conv_layer=1, agnostic_nms=False, outputNum=1, is_stream=False, norm=False):
+    if is_stream:
+        return run_image(image=im,src=src,inf_size=inf_size,obj_conf_thr=obj_conf_thr,iou_thr=iou_thr,conv_layer=conv_layer,agnostic_nms=agnostic_nms,outputNum=outputNum,is_stream=is_stream,norm=norm)
+    elif source_type == "Image":
+        return run_image(image=im,src=src,inf_size=inf_size,obj_conf_thr=obj_conf_thr,iou_thr=iou_thr,conv_layer=conv_layer,agnostic_nms=agnostic_nms,outputNum=outputNum,is_stream=is_stream,norm=norm)
+    elif source_type == "Video":
+        return run_video(video=vid,src=src,inf_size=inf_size,obj_conf_thr=obj_conf_thr,iou_thr=iou_thr,agnostic_nms=agnostic_nms,is_stream=is_stream,outputNum=outputNum)
+
+def run_image(image, src, inf_size, obj_conf_thr, iou_thr, conv_layer, agnostic_nms, outputNum, is_stream, norm):
+    """
+    Takes an image (from upload or webcam), and outputs the yolo7 boxed output and the convolution layers
+
+    Args:
+        image (str/PIL): The file path or PIL of the the input image.
+        src (str): The source of the input image, either upload or webcam
+        inf_size (int): The size of the inference
+        obj_conf_thr (float): The object confidence threshold
+        iou_thr (float): The intersection of union number
+        conv_layer (int): The number of the convolutional layer to show
+        agnostic_nms (bool): The agnostic nms boolean
+
+    Returns:
+        List[str]: A list of strings, where each string is a file path to an output image.
+    """
+    obj_conf_thr = float(obj_conf_thr)
+    iou_thr = float(iou_thr)
+    agnostic_nms = bool(agnostic_nms)
+    if src == "Webcam":
+        image.save('Temp.jpg')  # Convert PIL image to OpenCV format if needed
+        image = 'Temp.jpg'
+    if not is_stream:
+        random = Image.open(image)
+        new_dir = generate_feature_maps(random, conv_layer)
+    if agnostic_nms:
+        agnostic_nms = 'store_true'
+    else:
+        agnostic_nms = 'store_false'
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)')
+    parser.add_argument('--source', type=str, default=image, help='source')  # file/folder, 0 for webcam
+    parser.add_argument('--img-size', type=int, default=inf_size, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=obj_conf_thr, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=iou_thr, help='IOU threshold for NMS')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='display results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
+    parser.add_argument('--agnostic-nms', action=agnostic_nms, help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default='outputs/runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--no-trace', action='store_true', help='don`t trace model')
+    opt = parser.parse_args()
+    opt.no_trace = True
+    print(opt)
+    #check_requirements(exclude=('pycocotools', 'thop'))
+    if opt.update:  # update all models (to fix SourceChangeWarning)
+        for opt.weights in ['yolov7.pt']:
+            save_dir, smooth_dir, labels, formatted_time = detect(opt, outputNum=outputNum, is_stream=is_stream)
+            strip_optimizer(opt.weights)
+    else:
+        save_dir, smooth_dir, labels, formatted_time = detect(opt, outputNum=outputNum, is_stream=is_stream)
+    if is_stream:
+        return [save_dir, None, None, None, None, None]
+    return [save_dir, new_dir, smooth_dir, labels, formatted_time, None]  # added info
+
+def run_video(video, src, inf_size, obj_conf_thr, iou_thr, agnostic_nms, is_stream, outputNum=1, norm=False):
+    """
+    Takes a video (from upload or webcam), and outputs the yolo7 boxed output
+
+    Args:
+        video (str): The file path of the input video
+        src (str): The source of the input video, either upload or webcam
+        inf_size (int): The size of the inference
+        obj_conf_thr (float): The object confidence threshold
+        iou_thr (float): The intersection of union number
+        agnostic_nms (bool): The agnostic nms boolean
+
+    Returns:
+        str: The file path of the output video
+    """
+    obj_conf_thr = float(obj_conf_thr)
+    iou_thr = float(iou_thr)
+    agnostic_nms = bool(agnostic_nms)
+    if src == "Webcam":
+        if is_stream:
+            video = "0"
+        else:
+            video = correct_video(video)
+    if agnostic_nms:
+        agnostic_nms = 'store_true'
+    else:
+        agnostic_nms = 'store_false'
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)')
+    parser.add_argument('--source', type=str, default=video, help='source')  # file/folder, 0 for webcam
+    parser.add_argument('--img-size', type=int, default=inf_size, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=obj_conf_thr, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=iou_thr, help='IOU threshold for NMS')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='display results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
+    parser.add_argument('--agnostic-nms', action=agnostic_nms, help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default='outputs/runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--no-trace', action='store_true', help='don`t trace model')
+    opt = parser.parse_args()
+    opt.batch_size = 1
+    print(opt)
+    #check_requirements(exclude=('pycocotools', 'thop'))
+    with torch.no_grad():
+        if opt.update:  # update all models (to fix SourceChangeWarning)
+            for opt.weights in ['yolov7.pt']:
+                save_dir = detect(opt, outputNum=outputNum, is_stream=is_stream, norm=norm)
+                strip_optimizer(opt.weights)
+        else:
+            save_dir = detect(opt, outputNum=outputNum, is_stream=is_stream, norm=norm)
+    return [None, None, None, None, None, save_dir]

Interface_Dependencies/smooth_grad.py

@@ -0,0 +1,56 @@
+## Original Implementation of Smooth_grad **NOT USED**
+
+import torch
+import numpy as np
+
+def generate_vanilla_grad(model, input_tensor, outputNum, targets=None, norm=False, device='cpu'):
+    """
+    Generates an attribution map using vanilla gradient method.
+
+    Args:
+        model (torch.nn.Module): The PyTorch model to generate the attribution map for.
+        input_tensor (torch.Tensor): The input tensor to the model.
+        norm (bool, optional): Whether to normalize the attribution map. Defaults to False.
+        device (str, optional): The device to use for the computation. Defaults to 'cpu'.
+
+    Returns:
+        numpy.ndarray: The attribution map.
+    """
+    # Set requires_grad attribute of tensor. Important for computing gradients
+    input_tensor.requires_grad = True
+    
+    # Forward pass
+    train_out = model(input_tensor) # training outputs (no inference outputs in train mode)
+
+    num_classes = 2
+    
+    # Zero gradients
+    model.zero_grad()
+    
+    import torch
+
+    # train_out[1] = torch.Size([4, 3, 80, 80, 7]) #anchorxC) cls (class probabilities)
+    # train_out[0] = torch.Size([4, 3, 160, 160, 7]) #anchorx4) reg (location and scaling)
+    # train_out[2] = torch.Size([4, 3, 40, 40, 7]) #anchorx1) obj (objectness score or confidence)
+    
+    gradients = torch.autograd.grad(train_out[outputNum-1].requires_grad_(True), input_tensor, 
+                                    grad_outputs=torch.ones_like(train_out[outputNum-1]).requires_grad_(True), 
+                                    retain_graph=True, create_graph=True)
+    
+    # Convert gradients to numpy array
+    gradients = gradients[0].detach().cpu().numpy()
+
+    if norm:
+        # Take absolute values of gradients
+        gradients = np.absolute(gradients)
+
+        # Sum across color channels
+        attribution_map = np.sum(gradients, axis=0)
+
+        # Normalize attribution map
+        attribution_map /= np.max(attribution_map)
+    else:
+        # Sum across color channels
+        attribution_map = gradients
+    
+    return torch.tensor(attribution_map, dtype=torch.float32, device=device)

ffmpeg.7z

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b257499590550039b995ad1eb284ca5e5696c298bcd5cfd0aea325317ba62c1
+size 40956361

individual_work/asien/asien_interface.py

@@ -0,0 +1,29 @@
+import gradio as gr
+from PIL import Image
+import io
+import subprocess
+import os
+
+# Define a function to handle the image input
+def detect_objects(input_image):
+    # Save the uploaded image temporarily inside the "inference" folder
+    print(input_image)
+    
+    # Run your YOLOv7 detection script
+    subprocess.run(["python", r"yolov7-main\detect.py", "--source", input_image, "--project", "individual_work\\asien\\run_images", "--name", "exp"])
+
+
+    # Load the output image from your detection
+    output_image = Image.open("individual_work\\asien\\run_images\\exp\\image.png")
+    return output_image
+
+# Define the Gradio interface with a run button
+iface = gr.Interface(
+    fn=detect_objects,
+    inputs=gr.inputs.Image(type="filepath", source="upload"),
+    outputs=gr.outputs.Image(type="pil"),
+    live=False  # Set live=False to disable real-time updates
+)
+
+# Launch the Gradio interface
+iface.launch(share=True)

individual_work/braedon/braedon_settings.py

@@ -0,0 +1,69 @@
+import gradio as gr
+import argparse
+import sys
+sys.path.append('./')
+sys.path.append('yolov7-main')
+
+from ourDetect import detect
+import torch
+from utils.general import strip_optimizer
+
+# Define a function to run YOLOv7 with the provided settings
+def run(weights, conf_thres, iou_thres, agnostic_nms, source):
+    weights = weights.strip()  # Remove any leading/trailing spaces
+    conf_thres = float(conf_thres)
+    iou_thres = float(iou_thres)
+    agnostic_nms = bool(agnostic_nms)
+    source = source.strip()
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default=[weights], help='model.pt path(s)')
+    parser.add_argument('--source', type=str, default=source, help='source')
+    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=conf_thres, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=iou_thres, help='IOU threshold for NMS')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='display results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default='runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--no-trace', action='store_true', help='don`t trace model')
+    opt = parser.parse_args()
+    print(opt)
+
+    with torch.no_grad():
+        if opt.update:
+            for opt.weights in weights:
+                save_dir = detect(opt)
+                strip_optimizer(opt.weights)
+        else:
+            save_dir = detect(opt)
+    return save_dir + "\zidane.jpg"
+
+# Define the Gradio settings block
+settings_block = [
+    "text",  # "text" component for Weights (Path)
+    "number",  # "number" component for Confidence Threshold
+    "number",  # "number" component for IoU Threshold
+    "checkbox",  # "checkbox" component for Agnostic NMS
+    "text"  # "text" component for Source (Path)
+]
+
+# Create a Gradio interface for YOLOv7 settings
+iface = gr.Interface(
+    fn=run,
+    inputs=settings_block,
+    outputs="text",  # Use "text" directly as the output type
+    live=True
+)
+
+
+if __name__ == "__main__":
+    iface.launch()

individual_work/ike/olddetect.py

@@ -0,0 +1,248 @@
+import argparse
+import time
+from pathlib import Path
+
+import cv2
+import torch
+from PIL import Image
+import torch.backends.cudnn as cudnn
+import numpy as np
+from numpy import random
+
+import sys
+sys.path.append('./')
+sys.path.append('yolov7-main')
+
+
+from models.experimental import attempt_load
+from utils.datasets import LoadStreams, LoadImages
+from utils.general import check_img_size, check_requirements, check_imshow, non_max_suppression, apply_classifier, \
+    scale_coords, xyxy2xywh, strip_optimizer, set_logging, increment_path
+from utils.plots import plot_one_box
+from utils.torch_utils import select_device, load_classifier, time_synchronized, TracedModel
+
+
+def detect(input_image=None, input_Webcam=None):
+    source_img = None
+    save_txt = False
+    trace = False
+    # source = opt.source
+    if input_image:
+        source_img = np.array(input_image)  # Convert PIL image to OpenCV format if needed
+
+    if input_Webcam:
+        source_img = np.array(input_Webcam)  # Convert PIL image to OpenCV format if needed
+
+    if source_img is not None:
+        #source = cv2.cvtColor(cv2.imread(source), cv2.COLOR_RGB2BGR)
+
+        img = cv2.imdecode(np.fromstring(source_img(), np.uint8), 1)
+
+    # Convert image to YSBCR color space
+        source = cv2.cvtColor(img, cv2.COLOR_BGR2YCrCb)
+    else:
+        source, weights, view_img, save_txt, imgsz, trace = opt.source, opt.weights, opt.view_img, opt.save_txt, opt.img_size, not opt.no_trace
+        
+    # save_img = not opt.nosave and not source.endswith('.txt')  # save inference images
+    # webcam = source.isnumeric() or source.endswith('.txt') or source.lower().startswith(
+    #     ('rtsp://', 'rtmp://', 'http://', 'https://'))
+
+    # Directories
+    save_dir = Path(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))  # increment run
+    if not opt.nosave:
+        (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+    # Initialize
+    set_logging()
+    device = select_device(opt.device)
+    half = device.type != 'cpu'  # half precision only supported on CUDA
+
+    # Load model
+    weights = 'yolov7.pt'
+    imgsz = 640
+    model = attempt_load(weights, map_location=device)  # load FP32 model
+    stride = int(model.stride.max())  # model stride
+    imgsz = check_img_size(imgsz, s=stride)  # check img_size
+
+    # if trace:
+    #      model = TracedModel(model, device, opt.img_size)
+
+    if half:
+        model.half()  # to FP16
+
+    # Second-stage classifier
+    classify = False
+    if classify:
+        modelc = load_classifier(name='resnet101', n=2)  # initialize
+        modelc.load_state_dict(torch.load('weights/resnet101.pt', map_location=device)['model']).to(device).eval()
+
+    # Set Dataloader
+    dataset = LoadImages(source, img_size=imgsz, stride=stride)
+    view_img = check_imshow()
+    cudnn.benchmark = True
+    # if webcam:
+    #     view_img = check_imshow()
+    #     cudnn.benchmark = True  # set True to speed up constant image size inference
+    #     dataset = LoadStreams(source, img_size=imgsz, stride=stride)
+    # else:
+    #     dataset = LoadImages(source, img_size=imgsz, stride=stride)
+
+    # Get names and colors
+    names = model.module.names if hasattr(model, 'module') else model.names
+    colors = [[random.randint(0, 255) for _ in range(3)] for _ in names]
+
+    # Run inference
+    if device.type != 'cpu':
+        model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters())))  # run once
+    old_img_w = old_img_h = imgsz
+    old_img_b = 1
+
+    t0 = time.time()
+    for path, img, im0s, vid_cap in dataset:
+        img = torch.from_numpy(img).to(device)
+        img = img.half() if half else img.float()  # uint8 to fp16/32
+        img /= 255.0  # 0 - 255 to 0.0 - 1.0
+        if img.ndimension() == 3:
+            img = img.unsqueeze(0)
+
+        # Warmup
+        if device.type != 'cpu' and (old_img_b != img.shape[0] or old_img_h != img.shape[2] or old_img_w != img.shape[3]):
+            old_img_b = img.shape[0]
+            old_img_h = img.shape[2]
+            old_img_w = img.shape[3]
+            for i in range(3):
+                model(img, augment=opt.augment)[0]
+
+        # Inference
+        t1 = time_synchronized()
+        with torch.no_grad():   # Calculating gradients would cause a GPU memory leak
+            pred = model(img, augment=opt.augment)[0]
+        t2 = time_synchronized()
+
+        # Apply NMS
+        pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres, classes=opt.classes, agnostic=opt.agnostic_nms)
+        t3 = time_synchronized()
+
+        # Apply Classifier
+        if classify:
+            pred = apply_classifier(pred, modelc, img, im0s)
+
+        # Process detections
+        for i, det in enumerate(pred):  # detections per image
+            if input_Webcam:  # batch_size >= 1
+                p, s, im0, frame = path[i], '%g: ' % i, im0s[i].copy(), dataset.count
+            else:
+                p, s, im0, frame = path, '', im0s, getattr(dataset, 'frame', 0)
+
+            p = Path(p)  # to Path
+            save_path = str(save_dir / p.name)  # img.jpg
+            # txt_path = str(save_dir / 'labels' / p.stem) + ('' if dataset.mode == 'image' else f'_{frame}')  # img.txt
+            gn = torch.tensor(im0.shape)[[1, 0, 1, 0]]  # normalization gain whwh
+            if len(det):
+                # Rescale boxes from img_size to im0 size
+                det[:, :4] = scale_coords(img.shape[2:], det[:, :4], im0.shape).round()
+
+                # Print results
+                for c in det[:, -1].unique():
+                    n = (det[:, -1] == c).sum()  # detections per class
+                    s += f"{n} {names[int(c)]}{'s' * (n > 1)}, "  # add to string
+
+                # Write results
+                for *xyxy, conf, cls in reversed(det):
+                    # if save_txt:  # Write to file
+                    #     xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
+                    #     line = (cls, *xywh, conf) if opt.save_conf else (cls, *xywh)  # label format
+                    #     with open(txt_path + '.txt', 'a') as f:
+                    #         f.write(('%g ' * len(line)).rstrip() % line + '\n')
+
+                    if source or view_img:  # Add bbox to image
+                        label = f'{names[int(cls)]} {conf:.2f}'
+                        plot_one_box(xyxy, im0, label=label, color=colors[int(cls)], line_thickness=1)
+
+            # Print time (inference + NMS)
+            print(f'{s}Done. ({(1E3 * (t2 - t1)):.1f}ms) Inference, ({(1E3 * (t3 - t2)):.1f}ms) NMS')
+
+            # Stream results
+            if view_img:
+                cv2.imshow(str(p), im0)
+                cv2.waitKey(1)  # 1 millisecond
+
+            # Save results (image with detections)
+            if source:
+                if dataset.mode == 'image':
+                    if not opt.nosave:
+                        cv2.imwrite(save_path, im0)
+                    print(f" The image with the result is saved in: {save_path}")
+                else:  # 'video' or 'stream'
+                    if vid_path != save_path:  # new video
+                        vid_path = save_path
+                        if isinstance(vid_writer, cv2.VideoWriter):
+                            vid_writer.release()  # release previous video writer
+                        if vid_cap:  # video
+                            fps = vid_cap.get(cv2.CAP_PROP_FPS)
+                            w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+                            h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+                        else:  # stream
+                            fps, w, h = 30, im0.shape[1], im0.shape[0]
+                            save_path += '.mp4'
+                        vid_writer = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
+                    vid_writer.write(im0)
+
+    # if source:
+    #     s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+    #     #print(f"Results saved to {save_dir}{s}")
+
+    print(f'Done. ({time.time() - t0:.3f}s)')
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default='yolov7.pt', help='model.pt path(s)')
+    parser.add_argument('--source', type=str, default='inference/images', help='source')  # file/folder, 0 for webcam
+    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=0.25, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.45, help='IOU threshold for NMS')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--view-img', action='store_true', help='display results')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--nosave', action='store_true', help='do not save images/videos')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--project', default='runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--no-trace', action='store_true', help='don`t trace model')
+    opt = parser.parse_args()
+    print(opt)
+    #check_requirements(exclude=('pycocotools', 'thop'))
+
+    with torch.no_grad():
+        if opt.update:  # update all models (to fix SourceChangeWarning)
+            for opt.weights in ['yolov7.pt']:
+                detect()
+                strip_optimizer(opt.weights)
+        else:
+            detect()
+    import gradio as gr
+
+
+    
+    input_image = gr.inputs.Image(type='pil', label="Original Image", source="upload", optional=True)
+    input_Webcam = gr.inputs.Image(type='pil', label="Original Image", source="webcam", optional=True)
+    inputs = [input_image, input_Webcam]
+    outputs = gr.outputs.Image(type="pil", label="Output Image")
+    title = "Object detection with Yolov7"
+
+    iface = gr.Interface(detect(input_image, input_Webcam),
+                inputs = input_image, 
+                outputs = Image, 
+                title="Classification using YOLOV7",
+                live=True,
+                )
+    
+    iface.launch()
+
+    

references/error_fixes/online_help.md

@@ -0,0 +1,3 @@
+https://stackoverflow.com/questions/75103127/getting-notimplementederror-could-not-run-torchvisionnms-with-arguments-fr#:~:text=The%20full%20error%3A,(if%20using%20custom%20build).
+
+https://github.com/WongKinYiu/yolov7/issues/1205

references/gradio_exs/exstream.py

@@ -0,0 +1,13 @@
+import gradio as gr
+import numpy as np
+
+def flip(im):
+    return np.flipud(im)
+
+demo = gr.Interface(
+    flip, 
+    gr.Image(source='webcam', streaming=True), 
+    "image",
+    live=True
+)
+demo.launch()