tasks/detect/ #8489
Replies: 48 comments 150 replies
-
Will color differences in similar objects have an effect? For example, I created a model using black horse data and I will use it for white horses detection |
Beta Was this translation helpful? Give feedback.
-
I've finished trained my v8 detection model and i wanted to used a 3d camera (intel Realsense d435) to do inference on live data and then find the XYZ of the the centroid of the object boundary box in reference to the camera origin. How would i do this? |
Beta Was this translation helpful? Give feedback.
-
So is there a way to 'lock on' one particular detected object? Say I have the yolov8 detected 10 objects, my program then go through these objects and determine one particular object I want to do a 'lock' on it. Is there a way I can use yolov8 from that point to ignore the rest of the detected objects and just keep tracking/locking on the particular object I want? Thanks in advance. |
Beta Was this translation helpful? Give feedback.
-
It looks like I can not tell Yolov8 to stop other detections and just focus/locking on a particular detected object in real-time. Yes, it seems that a different tracking/locking algorithm needs to be in place for locking before I am done with it. I sure wish Yolov could have this ability to do so, I guess I can put this in a wish list.... Thanks for your reply. |
Beta Was this translation helpful? Give feedback.
-
Yes, thank you for your suggestion. I realized the tracking option in Yolov8 is there but still, it tracks multiple instead of just one. My program still needs filter logic to catch what I wanted. I am not sure how Yolov internal works but I presume excluding object by tracking ID will be much more faster than my python filtering logic...anyway, there will be a ton of applications can use the locking facility if provided, I sincerely wish you all can put this into you upcoming enhancements. Thanks for the reply. |
Beta Was this translation helpful? Give feedback.
-
Hey I am very much confused with the overall working of the YOLO model . I mean like when we say each cell is responsible for predicting an image what it actually means in terms of workflow and like what is the concept of anchor boxes like what do you mean by predefined anchor boxes. What i understand is that we have a image like (480,480,3) we finally convert it to like (7,7,30) by passing through convolutional layers / activation layers / pooiling layers . Assuming we divided it into 7 * 7 grid and for each cell we have a structure like lets say one anchor box x,y,w,h,conf,classess) but it is all prediction right ? we didnt defined anything so what does predefined anchor boxes mean? Also want to know one thing how are we suppossed to do transfer learning in object detection i dont get it . But how do we use transfer-learning in object detection i still don get it ? how is it divided into different stages ? Also in this doc for training yolov8 on custom dataset we are training all the layers which isnt transfer learning right and training all layers on a very small dataset wouldnt make sense right ? Can you please please please clear my confusion and share some updated/latest resources which could help me with this . I really need help in this. |
Beta Was this translation helpful? Give feedback.
-
Hi, for det in results[0].boxes.xyxy:
xmin, ymin, xmax, ymax, conf = map(int, det[:4].tolist() + [det[4]]) xyxy in 1,4 dimensional and therefor det[4] is out of bounds. what i did manage to get working but it feels convoluted since I am using [0] so many times for such a short code is from ultralytics import YOLO
import cv2
model = YOLO('./runs/detect/train/weights/best.pt')
result = model('tennis.jpg')
for result in result[0].boxes.numpy():
xyxy = result.xyxy[0]
x1, y1, x2, y2 = map(float, xyxy)
cls = int(result.cls[0])
conf = float(result.conf[0])
print("x1: ", x1, "y1: ", y1, "x2: ", x2, "y2: ", y2, "cls: ", cls, "conf: ", conf) |
Beta Was this translation helpful? Give feedback.
-
hi, im implementing yolov8 in my android gallery's search algorithm. the yolov8's purpose is to scan media files and return images that has a box label that matches the search query. i can make it work using yolov5s.torchscript.ptl with org.pytorch:pytorch_android_lite:1.10.0 and org.pytorch:pytorch_android_torchvision_lite:1.10.0 but i cant make it work with yolov8s.torchscript |
Beta Was this translation helpful? Give feedback.
-
Can you provide me with the best configurations for human-detection example using YOLOv8? |
Beta Was this translation helpful? Give feedback.
-
how to reduce predicted bounding box line width in yolov8 |
Beta Was this translation helpful? Give feedback.
-
Hi, can yolov8 detect color sequence?. i create color sequence of rainbow (red, yellow, green) as a real rainbow. if i change the yellow to orange, can yolov8 detect it as a wrong rainbow? |
Beta Was this translation helpful? Give feedback.
-
i need a object detection code using yolov8 for detect the two images |
Beta Was this translation helpful? Give feedback.
-
here is my code import cv2 Load the YOLO modelmodel = YOLO('yolov8n.pt') # Assuming 'yolov8n.pt' is the YOLO model you want to use Specify the path to your imageimage_path = '/media/XavierSSD250/22_08_23/ui.png' Perform object detectionresults = model(image_path) Convert image to OpenCV formatimage = cv2.imread(image_path) Extract bounding boxes, labels, and confidence scores from resultsbboxes = results.pred[0][:, :4] # Extract bounding boxes Draw bounding boxes and labels on the imagefor bbox, label, confidence in zip(bboxes, labels, confidences): Show the image with bounding boxescv2.imshow('Object Detection', image) |
Beta Was this translation helpful? Give feedback.
-
I am a beginner and have trained an object detection model using my own dataset. If I want to display the results of detection inference in real-time on the web, which method would be better if you recommend? |
Beta Was this translation helpful? Give feedback.
-
Hi, at the moment I'm trying to used .onnx exported from yolov8n.pt in Unity, using C#, to demonstrate real-time object detection in 3D scene, I got output in the shape of (1,84,8400). I would like to ask how to post process this output to draw bounding box, extracting the information of the output or the output format. Sorry for my english, and if there are any better method please suggest right away. thank you in advance. |
Beta Was this translation helpful? Give feedback.
-
Hi there, Thanks! |
Beta Was this translation helpful? Give feedback.
-
I already have a code for object detection in videos, now I want to include some evaluation metrics after running the model on the data. Is this possible? |
Beta Was this translation helpful? Give feedback.
-
Hi, The challenge is that the model does not detect anything on the validation set. The images are at least 4 megabytes. Any thoughts on this? My Code : from ultralytics import YOLO |
Beta Was this translation helpful? Give feedback.
-
Hey below is my code i want to apply nms algorithim to below code please help me. Please add NMS code to the below object detection code.I want to apply NMS since there are many boxes drawn around a single object.Please add NMS code to below code and please also check that code should be working.I earlier ask you then it give me error ultralytics do not have non_max_supperession. Below is the code. Thanks model = YOLO("runs/detect/train2/weights/last.pt")
|
Beta Was this translation helpful? Give feedback.
-
Good day, YOLO(yolov8n.pt) or YOLO(yolov8n.yaml) |
Beta Was this translation helpful? Give feedback.
-
Hey there! 🌟 |
Beta Was this translation helpful? Give feedback.
-
Hello and good day |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
Hey there, I am using yolo v8 to detect different brands in the shelves of retail stores. |
Beta Was this translation helpful? Give feedback.
-
Hi, is there a way i can filter out detection? i have a food detection model where i only want it to detect food that i trained on. However, as of now, it also detects non food items. How can i filter to only detect food? |
Beta Was this translation helpful? Give feedback.
-
I now have a trained yolov8 target detection model, which can detect 5 different types of targets. I now want to detect a new target and I have its corresponding training data. I want to do some fine-tuning on the model instead of retraining it, that is, transfer learning, so that my current model can detect targets for a total of 6 or even more models. How should I do it? Thanks |
Beta Was this translation helpful? Give feedback.
-
hello all I have around original images around 74 images dataset with resolution 8192 * 5460 Train the modelmodel.train( below is my results but the model is not performing well, in terms of precision. please give me hints how to imporve the model and how to make use of all the resources to reduce the training model also I have access to powerful machine with 5 GPU but I am not able to utilize all the resources the machine spec: NVIDIA-SMI 525.147.05 Driver Version: 525.147.05 CUDA Version: 12.0 | |
Beta Was this translation helpful? Give feedback.
-
How to disable print outputs like this: “0: 384x640 3 bags, 124.8ms |
Beta Was this translation helpful? Give feedback.
-
Hello, I want to improve the detection accuracy of specific classes in the dataset without using retraining. My current idea is to increase the weight of specific classes in the loss function. Is this practice effective? Are there other effective ways to do this? |
Beta Was this translation helpful? Give feedback.
-
tasks/detect/
Official documentation for YOLOv8 by Ultralytics. Learn how to train, validate, predict and export models in various formats. Including detailed performance stats.
https://docs.ultralytics.com/tasks/detect/
Beta Was this translation helpful? Give feedback.
All reactions