Object Tracking using YOLOv8 on Custom Dataset
Вставка
- Опубліковано 26 січ 2023
- Learn to track custom objects using YoloV8 and different Object Trackers.
The detections generated by YOLOv8, a family of object detection architectures and models pretrained on the Boat dataset, are passed to the tracker of your choice. Supported ones at the moment are: StrongSORT OSNet, OCSORT and ByteTrack. They can track any object that your Yolov8 model was trained to detect.
#####################################################################
For queries: Comment in comment section or you can mail me at aarohisingla1987@gmail.com
#####################################################################
#objectdetection #objecttracking #pytorch #python - Наука та технологія
Thank you for your great video @ AArohi
What an amazing video.. Worth watching. It again & again
Glad you enjoyed it
Thank you so much for sharing this video. It helped me a Lot. Kindly keep sharing ma`am
Thank you, I will
Best video of VOLOv8
Thank you!
Thank you so much mam! i have been waiting for this video for a long time
I will upload the colab video tomorrow :)
@@CodeWithAarohi okay thank you mam
VEry nicely explained
Thank you so much 🙂
Is the data annotated seperately using annotation tools since it seems boat_yacht_dataset folder and dataset folder with train and valid(with labels and images) and simply train and val folders respectively.
thank you so much
Well done
Thanks!
Eagerly waiting for your video on using this yolov8 micel in drone to detect object real-time & send notification to db or other Kafka consumer !
I will try my best
Well done!!. It’d be nice to incorporate counting for one class and multiple ones and saving results in csv, this is one of the challenge thing in CV
Thankyou for the suggestion. Will try to cover the suggestion in upcoming videos 🙂
Thank You for sharing this, can we draw tracks too using this code?
Hello arohi, how we can use custom model.weight in production as i am working with yolov5 do we need clone whole official repo ?
I am using yolo v8 , with bytetrck , some time track Id change,could you please share configuration of bytetrack
Thank you for this great video on how to prepare a dataset for feature extractor model! Is this the same way for ReID?
Yes, it is
terima kasih ilmunya
Glad to help!
Thank you very much for these explanations, How can we organize our dataset for feature extration models if we have two classes for example?
Emailed 🙂
Madam pls explain semi supervised learning techniques on yolov8
Thank you so much mam
Most welcome 😊
How can we use the yolo in the extremly large dataset i.e (like having 1 lakh images with more objects) ? Wouldnt it will be difficult to manually anotate the larger set..?
Yes, It will be a time consuming process but you have to annotate the whole dataset. There are organizations who provide paid annotation services.
I am getting unknown model error it mentions the model must be from resnet18,resnet34,resnet50…etc I have built my own weights for yolo weights and Reid weights kindly help me to solve this
mam can you apply this code on person detection and tracking
Hello Maam, I am trying to use Yolo V8 for object detection through camera and along with that i want coordinates of the bounding boxes. Can you please guide me how can i do that?
You can extract the coordinates of the bounding boxes by parsing the output of the YOLOv8 model.
Can you please tell me how to calculate the overall accuracy(SHOWING mAP score) of the particular trained yolov7 model in Google colab?
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]
Madam how to do the feature extraction of a dataset which has multiple classes.(Like Cats & Dogs). Here only having a single class which is boat class.
Hello, did you have an answer to your question?
I am training object tracking model on visdrone dataset. I followed exactly same approach as mentioned in video. But while performing visualization in "visualize_model" function, it is giving error: `Invalid shape (3, 224, 224) for image data` in line: `plt.imshow(inputs.cpu().data[j])`. also the model so trained is getting rejected while performing the actual tracking..(saying it is not of resnet18 type).. Please let me know where I may be wrong..Thank you mam for providing a good resource in domain of object tracking.
@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!!
can i use json file as input?
hi! is your jupyter notebook available somewhere?
Sorry, no
can we also use that feature extractor to recognize humans individually?
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
mam did you upload it to your github repo
????
best_model_wts = copy.deepcopy(model.state_dict())---- this line takes infinite time to execute, pl help to reduce the time. Is it because of HD quality of images?
Reduce image size and batch size
what is difference obect detection and object tracking
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.
I have a problem where it says graph execution error..could you make a video on this?
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"
I have a doubt I trained my model but how to count number of object of that label
Will try to cover in upcoming videos
@@CodeWithAarohi please just try to explain or make as soon as possible it's urgent project 😅
Hi Ma'am, Can you please share the code
i couldn't find the guide in pytorch docs. can you share the code?
pytorch.org/tutorials/beginner/introyt/modelsyt_tutorial.html
Please share the link of this feature extractor on pytorch official website
pytorch.org/vision/stable/feature_extraction.html
How to use custom dataset on real-time project? Can anyone guide me
The process I have explained in this video is for custom dataset.
Can you please provide collab liink?
Mam i need the dataset from this video can i get it from u please ..please reply mam
Emailed
Thanks!
Welcome!
How predict movement of object
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
can you please share the code
Dataset kahan hai ma'am. plz provide.
You have to prepare dataset yourself.
Where is the Source Code or Git Repo Link Ma'am ?
This code is not posted on Github.
Can you share your folder link please🥹🥹
This code is available for channel members (Contribution level 2)