Object Tracking using YOLOv8 on Custom Dataset 

Thank you, I will

Yes, it is

Thank you so much 🙂

This is available for channel members (Contribution level 2)

Reduce image size and batch size

Yes, It will be a time consuming process but you have to annotate the whole dataset. There are organizations who provide paid annotation services.

Hello, did you have an answer to your question?

pytorch.org/vision/stable/feature_extraction.html

@CodeWithAarohi I think you need to use plt.imshow instead of only imshow since it will give error. Kindly do correct me if I'm wrong.

If plt.imshow is working then use it. Otherwise as per your error: the error message indicates that the shape of the input tensor is (3, 224, 224). The first dimension (3) likely corresponds to the number of color channels in the image (e.g., red, green, blue), while the next two dimensions (224, 224) correspond to the height and width of the image. To fix this error, you may need to reshape your input data to match the expected shape of the plt.imshow() function. One way to do this is to use the transpose() function in PyTorch to change the order of the dimensions in your tensor: # Assume that "inputs" is your input tensor with shape (3, 224, 224) inputs = inputs.transpose(0, 2).transpose(0, 1) # Transpose the dimensions to (224, 224, 3) plt.imshow(inputs.cpu().data[j]) # Display the image

@@CodeWithAarohi I use plt.imshow(inputs.cpu().data[j].T) # Display the image to make dimensions as (224, 224, 3) but now in the output the images are like film negatives which are not relevant to val dataset

@Code With Aarohi, please share link of this notebook. I will cross check the code. Maybe I am wrong somewhere. Thank you!!

The process I have explained in this video is for custom dataset.

????

Detection involves placing a bounding box around an object and tracking means determining whether that object appears in next frames and for how long, by assigning a unique ID to each object.

This code is available for channel members (Contribution level 2)

Yes, feature extractors can indeed be used for recognizing humans individually

@@CodeWithAarohi is this folders and code available in github by chance. bc its very important for my FYP research.🙂😊

@@CodeWithAarohi is that code and folders available on GitHub by chance ?this project is very important for my FYP researches. thank you very much

Emailed 🙂

You can extract the coordinates of the bounding boxes by parsing the output of the YOLOv8 model.

from utils.metrics import ap_per_class # define the confidence threshold and IoU threshold for non-maximum suppression conf_thres = 0.25 iou_thres = 0.45 # define the test set of images test_images = ['path/to/your/test/images'] # define the number of classes in your dataset num_classes = 1 # generate predictions on the test set prediction = [] targets = [] for path, img, im0s, _ in LoadImages(test_images, img_size=640): img = torch.from_numpy(img).to(model.device).float() img /= 255.0 if img.ndimension() == 3: img = img.unsqueeze(0) with torch.no_grad(): prediction += model(img)[0] targets += LoadImages.detect(exists=True)[0][1] # calculate the AP per class metrics_output = ap_per_class(prediction, targets, plot=False, fname="precision-recall_curve.png", imgsz=imgsz) mAP = metrics_output['mAP'][0]

I will try my best

Emailed

pytorch.org/tutorials/beginner/introyt/modelsyt_tutorial.html

Sorry, no

This code is not posted on Github.

Thankyou for the suggestion. Will try to cover the suggestion in upcoming videos 🙂

Pleas explain the error in more detailed way. I will try to help

@@CodeWithAarohi mam I have a football dataset and I'm predicting the winner for a betting bot,I'm using the LSTM model ...on the first epoch it gives the error says "graph execution error"

Will try to cover in upcoming videos

@@CodeWithAarohi please just try to explain or make as soon as possible it's urgent project 😅

I will upload the colab video tomorrow :)

@@CodeWithAarohi okay thank you mam

You have to prepare dataset yourself.

Glad you enjoyed it

Please elaborate your question

@@CodeWithAarohi i want to identify object for example man then i want to predict that man go to right or left

Most welcome 😊

Glad to help!

Thank you!

Thanks!

Welcome!