Adding polygon detection alongside bounding boxes #6506

ahmad4633 · 2022-02-02T13:06:34Z

I would like to thank XinzeLee for his amazing work, that inspired this pull request.

This PR adds 4 points polygon detection capability to this repo. Original bounding box detection is not modified. The new head will output 9 values (index x1 y1 x2 y2 x3 y3 x4 y4), the labels for training are of the same format as the mentioned output.

Also, for training we need to install polygon IoU Cuda to speed up the validation process and we need set "--polygon True", a full tutorial for training is provided in this notebook Polygon Tutorial.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced support for polygon labels in YOLOv5.

📊 Key Changes

Added polygon label plotting utilities.
Improved polygon label support in training and logging functions.
Integrated polygon-specific NMS (Non-Maximum Suppression).
Developed CUDA extensions for polygon IOU (Intersection Over Union) computation.

🎯 Purpose & Impact

These changes allow YOLOv5 to train and infer with polygon-shaped bounding boxes, which are more accurate for certain objects.
The CUDA extensions improve performance by offloading complex IOU calculations to the GPU.
Overall, this broadens the range of computer vision tasks YOLOv5 can handle effectively.

and polygon yolo

for more information, see https://pre-commit.ci

github-actions

👋 Hello @ahmad4633, thank you for submitting a YOLOv5 🚀 PR! To allow your work to be integrated as seamlessly as possible, we advise you to:

✅ Verify your PR is up-to-date with upstream/master. If your PR is behind upstream/master an automatic GitHub Actions merge may be attempted by writing /rebase in a new comment, or by running the following code, replacing 'feature' with the name of your local branch:

git remote add upstream https://github.com/ultralytics/yolov5.git
git fetch upstream
# git checkout feature  # <--- replace 'feature' with local branch name
git merge upstream/master
git push -u origin -f

✅ Verify all Continuous Integration (CI) checks are passing.
✅ Reduce changes to the absolute minimum required for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." -Bruce Lee

for more information, see https://pre-commit.ci

remove static path

for more information, see https://pre-commit.ci

… into test/polygon_yolo

for more information, see https://pre-commit.ci

… into test/polygon_yolo

ahmad4633 · 2022-02-15T14:42:55Z

I keep getting load_image is not defined error even though it should be imported from utils/datasets ? Any help ? Also I have installed iou_cuda utils via setup.py but I am getting Warning: "polygon_inter_union_cuda" and "polygon_b_inter_union_cuda" are not installed errors ?

Merhaba Uygar, please refer to Colab Tutorial.

for more information, see https://pre-commit.ci

tujh2 · 2022-02-15T17:28:05Z

Hi, @RoudyES

Unfortunately, there is no direct fix from our end for this.

I figured out that you can just comment y[..., 8:] = y[..., 8:].sigmoid() line in Polygon_Detect.forward(x) in models/yolo.py and then OpenCV reads model successfully. And it does not fail on Mul operator too.
The graph after exporting does not contain ScatterND operators.

Also I have a question about Polygon_Detect._make_grid() func
OpenCV fails on forward() step. I've got the same error as in opencv/opencv#20072
I just copied original _make_grid() func from Detect and this helped.
Now I can use polygon model with OpenCV and this working fine (I have custom inference pipeline in c++ and opencv, so I don't use detect.py)

Also I don't know if this changes are correct, but I'll leave that question to you 🤷

RoudyES · 2022-02-15T18:15:21Z

@tujh2 Careful of removing that y[..., 8:] = y[..., 8:].sigmoid(), it is there to make sure that all the outputs are within the [0..1] range. Without it the model might output values outside of that range which will go out of bounds of the image. You should be able to remedy that in post processing but we thought of leaving it there to include that step directly in the model and save the user from an extra step at post processing. I think it is also present in the master branch for the same reason (clamping values between 0..1), @glenn-jocher can correct me if I'm wrong.

As for the _make_grid() function I'm glad that it worked fine for you! Thank you for pointing that swapping it with Detect's function fixed it! We'll do some further testing to make sure that swapping the functions won't have any hidden unwanted effect, if all goes well we'll then update the PR with the original _make_grid().

Also, if you're on C++ I recommend swapping your inference engine to TensorRT, your pre and post processing steps will remain almost the same and you'll get a significant speedup in your pipeline.

sctrueew · 2022-02-15T19:02:38Z

@RoudyES Hi,

I've converted the model to onnx and when I want to do inference in C++ I get this error:

(-2:Unspecified error) Can't create layer "416" of type "Range" in function 'cv::dnn::dnn4_v20211004::LayerData::getLayerInstance'

Do you have a sample to use in C++ on onnx or TensorRT?

Thanks

RoudyES · 2022-02-15T19:32:01Z

@sctrueew it seems to me that you tried to run inference using OpenCV's DNN module. If so, please refer to @tujh2's comment above. They made some slight changes and were able to run the model with OpenCV's DNN.

However, I recommend running ONNX models using ONNX Runtime. Also, if you compile onnx runtime from source with TensorRT execution provider you will be able to run the same .onnx model with TensorRT backend without changing your code (the runtime will automatically convert the onnx model to .engine). Or you can just export the .pt model into .engine using export.py.

As for the sample, I'm afraid we don't have a C++ snippet. We usually run everything in C#.

sctrueew · 2022-02-15T20:02:11Z

@RoudyES Thank you. @tujh2, Have you successfully run in C++?

tujh2 · 2022-02-15T20:26:32Z

@sctrueew
Yes, I have. I run it with OpenCV 4.5.4 successfully.
I think that you exported your model without --simplify option. Try with this option and my tips in comment above.

sctrueew · 2022-02-15T20:28:03Z

@sctrueew it seems to me that you tried to run inference using OpenCV's DNN module. If so, please refer to @tujh2's comment above. They made some slight changes and were able to run the model with OpenCV's DNN.

However, I recommend running ONNX models using ONNX Runtime. Also, if you compile onnx runtime from source with TensorRT execution provider you will be able to run the same .onnx model with TensorRT backend without changing your code (the runtime will automatically convert the onnx model to .engine). Or you can just export the .pt model into .engine using export.py.

As for the sample, I'm afraid we don't have a C++ snippet. We usually run everything in C#.

@RoudyES Hi,
Thank you for the reply, My final goal is to run it in C# as well. It would be nice to have it in C#.

sctrueew · 2022-02-15T20:53:28Z

@sctrueew Yes, I have. I run it with OpenCV 4.5.4 successfully. I think that you exported your model without --simplify option. Try with this option and my tips in comment above.

Thanks, I've exported successfully the model to onnx but I still get the error when I want to readNet.

OpenCV version: 4.5.4 => built with GPU
auto net = dnn::readNetFromONNX("polygon.onnx");

What changes did you in the C++ side?

tujh2 · 2022-02-15T21:20:55Z

What changes did you in the C++ side?

Nothing at all. Actually I use GoCV - golang bindings for OpenCV, so I didn't modify code inside OpenCV library at all

sctrueew · 2022-02-15T21:49:15Z

What changes did you in the C++ side?

Nothing at all. Actually I use GoCV - golang bindings for OpenCV, so I didn't modify code inside OpenCV library at all

I've followed your tips and finally, I could readNet but I get an error in forward():

shape_utils.hpp:171: error: (-215:Assertion failed) start <= (int)shape.size() && end <= (int)shape.size() && start <= end in function 'cv::dnn::dnn4_v20211004::total'

tujh2 · 2022-02-15T21:54:03Z

What changes did you in the C++ side?

Nothing at all. Actually I use GoCV - golang bindings for OpenCV, so I didn't modify code inside OpenCV library at all

I've followed your tips and finally, I could readNet but I get an error in forward():

shape_utils.hpp:171: error: (-215:Assertion failed) start <= (int)shape.size() && end <= (int)shape.size() && start <= end in function 'cv::dnn::dnn4_v20211004::total'

This is a problem with Polygon_Detect._make_grid() func.

    def forward(self, x):
        # x = x.copy() # for profiling
        z = []  # inference output
        for i in range(self.nl):
            x[i] = self.m[i](x[i])  # conv
            bs, _, ny, nx = x[i].shape  # x(bs,267,20,20) to x(bs,3,20,20,89)
            x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()

            if not self.training:  # inference
                if self.onnx_dynamic or self.grid[i].shape[2:4] != x[i].shape[2:4]:
                    self.grid[i] = self._make_grid(nx, ny, i)

                y = x[i].clone()
                #y[..., 8:] = y[..., 8:].sigmoid()
                # y[..., 8], y[..., 9:] = y[..., 8].sigmoid(), y[...,
                # 9:].softmax(dim=-1) # softmax loss for classes
                if self.inplace:
                    y[..., :8] = (y[..., :8] + self.grid[i].repeat((1, 1, 1, 1, 4))) * self.stride[i]  # xyxyxyxy
                else:
                    xyxyxyxy = (y[..., :8] + self.grid[i].repeat((1, 1, 1, 1, 4))) * self.stride[i]  # xyxyxyxy
                    y = torch.cat((xyxyxyxy, y[..., 8:]), -1)
                z.append(y.view(bs, -1, self.no))

        return x if self.training else (torch.cat(z, 1), x)

    def _make_grid(self, nx=20, ny=20, i=0):
        d = self.anchors[i].device
        if check_version(torch.__version__, '1.10.0'):  # torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibility
            yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)], indexing='ij')
        else:
            yv, xv = torch.meshgrid([torch.arange(ny, device=d), torch.arange(nx, device=d)])
        return torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2)).float()

Try replace code in Polygon_Detect with this. That's work for me.

sctrueew · 2022-02-15T22:09:40Z

@tujh2 Thank you very much. it works for me but I have to change the result of forwarding.

`
cv::dnn::blobFromImage(input_image, blob, 1. / 255., cv::Size(INPUT_WIDTH, INPUT_HEIGHT), cv::Scalar(), true, false);

net.setInput(blob);
std::vector<cv::Mat> outputs;
net.forward(outputs, net.getUnconnectedOutLayersNames());

float x_factor = input_image.cols / INPUT_WIDTH;
float y_factor = input_image.rows / INPUT_HEIGHT;

float* data = (float*)outputs[0].data;

const int dimensions = 85;
const int rows = 25200;

std::vector<int> class_ids;
std::vector<float> confidences;
std::vector<cv::Rect> boxes;

for (int i = 0; i < rows; ++i) {
	float confidence = data[4];`  <<===== error occurred`

sctrueew · 2022-02-15T22:55:36Z

@tujh2 Do you have any ideas on how to resolve this issue?
Thanks

tujh2 · 2022-02-15T23:17:33Z

@tujh2 Do you have any ideas on how to resolve this issue? Thanks

I can publish my Go-inference code. Just modify this for C++.

        blob := gocv.BlobFromImage(src, 1/255.0, myNet.NetSize, myNet.netMean, true, false)
        defer blob.Close()
	myNet.net.SetInput(blob, "")

	netOutput := myNet.net.Forward(myNet.outputLayerName) // outputLayerName = "output"
	defer netOutput.Close()
	srcImgSize := src.Size()
	ratioW := float32(srcImgSize[1]) / float32(myNet.NetSize.Y) // NetSize is like --img (default 640x640)
	ratioH := float32(srcImgSize[0]) / float32(myNet.NetSize.X)
	var netWidth int = myNet.СlassesCount
	if myNet.isPolygon {
		netWidth += 9 // polygon has 8 fields for coords and 1 field for boxScore and other - class score
	} else {
		netWidth += 5
	}
	pData, _ := netOutput.DataPtrFloat32()
	var rects []image.Rectangle
	var polygonRects []PolygonOutput
	var scores []float32
	var indices []int
	var classes []int
	for i := 0; i < len(pData)-netWidth; i += netWidth {
		if myNet.isPolygon {
			fboxScore := pData[i+8]
			if fboxScore > myNet.BoxThreshold {
				x1 := pData[i] * ratioW
				y1 := pData[i+1] * ratioH
				x2 := pData[i+2] * ratioW
				y2 := pData[i+3] * ratioH
				x3 := pData[i+4] * ratioW
				y3 := pData[i+5] * ratioH
				x4 := pData[i+6] * ratioW
				y4 := pData[i+7] * ratioH

				var maxClassScore float32 = 0.0
				var classIndex = 0
				for k := i + 9; k < i+9+myNet.СlassesCount; k++ {
					classScore := pData[k]
					if classScore > maxClassScore {
						classIndex = k
						maxClassScore = classScore
					}
				}
			        // save result
			        // classId := classIndex - i - 9
			        //...
			}
		} else { 
		        // non-polygon yolo
		        //...
		}
	}

sctrueew · 2022-02-16T05:39:31Z

@tujh2 Do you have any ideas on how to resolve this issue? Thanks

I can publish my Go-inference code. Just modify this for C++.

        blob := gocv.BlobFromImage(src, 1/255.0, myNet.NetSize, myNet.netMean, true, false)
        defer blob.Close()
	myNet.net.SetInput(blob, "")

	netOutput := myNet.net.Forward(myNet.outputLayerName) // outputLayerName = "output"
	defer netOutput.Close()
	srcImgSize := src.Size()
	ratioW := float32(srcImgSize[1]) / float32(myNet.NetSize.Y) // NetSize is like --img (default 640x640)
	ratioH := float32(srcImgSize[0]) / float32(myNet.NetSize.X)
	var netWidth int = myNet.СlassesCount
	if myNet.isPolygon {
		netWidth += 9 // polygon has 8 fields for coords and 1 field for boxScore and other - class score
	} else {
		netWidth += 5
	}
	pData, _ := netOutput.DataPtrFloat32()
	var rects []image.Rectangle
	var polygonRects []PolygonOutput
	var scores []float32
	var indices []int
	var classes []int
	for i := 0; i < len(pData)-netWidth; i += netWidth {
		if myNet.isPolygon {
			fboxScore := pData[i+8]
			if fboxScore > myNet.BoxThreshold {
				x1 := pData[i] * ratioW
				y1 := pData[i+1] * ratioH
				x2 := pData[i+2] * ratioW
				y2 := pData[i+3] * ratioH
				x3 := pData[i+4] * ratioW
				y3 := pData[i+5] * ratioH
				x4 := pData[i+6] * ratioW
				y4 := pData[i+7] * ratioH

				var maxClassScore float32 = 0.0
				var classIndex = 0
				for k := i + 9; k < i+9+myNet.СlassesCount; k++ {
					classScore := pData[k]
					if classScore > maxClassScore {
						classIndex = k
						maxClassScore = classScore
					}
				}
			        // save result
			        // classId := classIndex - i - 9
			        //...
			}
		} else { 
		        // non-polygon yolo
		        //...
		}
	}

Thanks for sharing, If we have points more than 4 points how can we set (netWidth += 9). I don't know how many points are there? for example, one detection has 4 (x,y) and another one has 12(x,y)

sctrueew · 2022-02-16T07:01:20Z

This is my code but not working.
`

	Mat blob;
        int col = SrcImg.cols;
	int row = SrcImg.rows;
	int maxLen = MAX(col, row);
	Mat netInputImg = SrcImg.clone();
	if (maxLen > 1.2 * col || maxLen > 1.2 * row) {
		Mat resizeImg = Mat::zeros(maxLen, maxLen, CV_8UC3);
		SrcImg.copyTo(resizeImg(Rect(0, 0, col, row)));
		netInputImg = resizeImg;
	}
	blobFromImage(netInputImg, blob, 1 / 255.0, cv::Size(netWidth, netHeight), cv::Scalar(104, 117, 123), true, false);
	net.setInput(blob);
	std::vector<cv::Mat> netOutputImg;
	net.forward(netOutputImg, net.getUnconnectedOutLayersNames());
	std::vector<int> classIds;
	std::vector<float> confidences;
	std::vector<cv::Rect> boxes;
	float ratio_h = (float)netInputImg.rows / netHeight;
	float ratio_w = (float)netInputImg.cols / netWidth;
	//int net_width = className.size() + 5;
	int net_width = className.size() + 9;
	float* pdata = (float*)netOutputImg[0].data;



	std::vector<polygon> pol;

	float classScore = 0;
	for (int i = 0; i < sizeof(pdata) - net_width; i += net_width) {

		float confidence = pdata[i + 8];
		if (confidence >= 0.5) {

			float x1 = pdata[i] * ratio_w;
			float y1 = pdata[i + 1] * ratio_h;
			float x2 = pdata[i + 2] * ratio_w;
			float y2 = pdata[i + 3] * ratio_h;
			float x3 = pdata[i + 4] * ratio_w;
			float y3 = pdata[i + 5] * ratio_h;
			float x4 = pdata[i + 6] * ratio_w;
			float y4 = pdata[i + 7] * ratio_h;
			auto maxClassScore = 0.0;
			auto classIndex = 0;
			for (int k = i + 9; k < i + 9 + className.size(); k++) {
				classScore = pdata[k];
				if (classScore > maxClassScore) {
					classIndex = k;
					maxClassScore = classScore;
				}
			}
			classIds.push_back(classIndex);
			confidences.push_back(maxClassScore);
			//boxes.push_back(Rect(x1, y1, x2, y2));
			pol.push_back({ x1,y1,x2,y2,x3,y3,x4,y4 });
		}
	}

`

RoudyES · 2022-02-16T07:25:44Z

@sctrueew Please note that the model can only predict 4 points polygons. You will never have a result with more than 4 points (if you did your post processing isn't correct).

sctrueew · 2022-02-16T08:30:57Z

Yes, I know. what about this example?

There are more than four points for each object in the example above

RoudyES · 2022-02-16T08:37:11Z

@sctrueew This can't be achieved in this PR. The repo can only detect bounding boxes (such as the rotated boxes of the persons in this example). The image you shared seems to me the result of an Instance Segmentation model, where you get an object's bounding box along with it's contour polygon, this is not yet supported in this PR (although we might look into it in the future if this PR gets merged with master branch).

GivanTsai · 2022-02-18T06:24:20Z

There is a bug when I want to save the predicted text. If I run the inference command
python detect.py --polygon True --data '/content/yolov5/yolov5/data/polygon.ymal' --weights '/content/gdrive/MyDrive/Polygon_Yolo/yolov5/lpr/s/exp/weights/polygon_best.pt' --img 320 --source '/content/yolov5/polygon_all/images/val/' --hide-labels
It is Successful, but it is failed when I add the option --save-txt and --save-crop
!python detect.py --polygon True --data '/content/yolov5/yolov5/data/polygon.ymal' --weights '/content/gdrive/MyDrive/Polygon_Yolo/yolov5/lpr/s/exp/weights/polygon_best.pt' --img 320 --source '/content/yolov5/polygon_all/images/val/' --hide-labels --save-txt --save-conf --save-crop

How to deal with this bug? @ahmad4633

GivanTsai · 2022-02-18T06:42:37Z

There is a bug when I want to save the predicted text. If I run the inference command python detect.py --polygon True --data '/content/yolov5/yolov5/data/polygon.ymal' --weights '/content/gdrive/MyDrive/Polygon_Yolo/yolov5/lpr/s/exp/weights/polygon_best.pt' --img 320 --source '/content/yolov5/polygon_all/images/val/' --hide-labels It is Successful, but it is failed when I add the option --save-txt and --save-crop !python detect.py --polygon True --data '/content/yolov5/yolov5/data/polygon.ymal' --weights '/content/gdrive/MyDrive/Polygon_Yolo/yolov5/lpr/s/exp/weights/polygon_best.pt' --img 320 --source '/content/yolov5/polygon_all/images/val/' --hide-labels --save-txt --save-conf --save-crop How to deal with this bug? @ahmad4633

It seems that the size of gn is torch.Size([4]) and torch.tensor(xyxyxyxy).view(1, 8) is torch.Size([1,8]), how to fix it?

github-actions · 2022-03-21T00:18:12Z

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions YOLOv5 🚀 and Vision AI ⭐.

Aun0124 · 2023-01-11T08:04:51Z

@sctrueew @tujh2 I have also convert my model to onnx and the output shape will be (batch, x, 12) for polygon . Can i know did you all implement own nms or using other solution to solve this ? Anyway I wanted to implement in java for android apps.

tujh2 · 2023-01-11T09:28:31Z

@sctrueew @tujh2 I have also convert my model to onnx and the output shape will be (batch, x, 12) for polygon . Can i know did you all implement own nms or using other solution to solve this ? Anyway I wanted to implement in java for android apps.

Hi, no, i have missed nms step cause my photo context contains only 1 object. So i just choose polygon by max score. Also, as i know, polygon nms is an expensive step

Aun0124 · 2023-01-16T02:58:11Z

@sctrueew @tujh2 I have also convert my model to onnx and the output shape will be (batch, x, 12) for polygon . Can i know did you all implement own nms or using other solution to solve this ? Anyway I wanted to implement in java for android apps.

Hi, no, i have missed nms step cause my photo context contains only 1 object. So i just choose polygon by max score. Also, as i know, polygon nms is an expensive step

@tujh2 how about the preprocessing letterbox resize? can i have look on your code anyway , will be greatful.

ata and others added 15 commits January 25, 2022 16:29

adding polygon yolo

202e62e

last changes

ca9d901

more cleaning

6ad3f30

added polygon_train bool to parse model

2dd00f1

modifying detect for yolo and polygon

2d61899

polygon plot results

9e8f8ed

dataset fix

1828e48

merging latest yolo commits

6fb367d

and polygon yolo

Delete lpr.yaml

2754448

cleaning

6be048b

last changes

218a731

Merge remote-tracking branch 'origin/master' into test/polygon_yolo

38d936c

fixing polygon_create_dataloader

3e16e13

fixing plot_results

3e974f2

[pre-commit.ci] auto fixes from pre-commit.com hooks

bb7026a

for more information, see https://pre-commit.ci

github-actions bot reviewed Feb 2, 2022

View reviewed changes

ahmad4633 and others added 3 commits February 2, 2022 15:19

Update requirements.txt

ce3527b

[pre-commit.ci] auto fixes from pre-commit.com hooks

fc6aa18

for more information, see https://pre-commit.ci

Update detect.py

6f10192

remove static path

ahmad4633 closed this Feb 2, 2022

ahmad4633 reopened this Feb 3, 2022

ahmad4633 and others added 6 commits February 4, 2022 09:46

fixing indents

183e2d3

Merge remote-tracking branch 'origin/master' into test/polygon_yolo

66a53f0

[pre-commit.ci] auto fixes from pre-commit.com hooks

3f30d43

for more information, see https://pre-commit.ci

pre-commit.ci

7d077d5

Merge branch 'test/polygon_yolo' of https://github.com/ahmad4633/yolov5…

b0282f9

… into test/polygon_yolo

[pre-commit.ci] auto fixes from pre-commit.com hooks

7de5778

for more information, see https://pre-commit.ci

glenn-jocher assigned ahmad4633 Feb 4, 2022

ahmad4633 added 2 commits February 5, 2022 08:27

solving conflicts

bdb0994

Merge branch 'test/polygon_yolo' of https://github.com/ahmad4633/yolov5…

03235fb

… into test/polygon_yolo

ahmad4633 and others added 2 commits February 15, 2022 16:47

adding polygon_load_image

5cc70b6

[pre-commit.ci] auto fixes from pre-commit.com hooks

efb59a4

for more information, see https://pre-commit.ci

github-actions bot added the Stale label Mar 21, 2022

github-actions bot closed this Mar 27, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adding polygon detection alongside bounding boxes #6506

Adding polygon detection alongside bounding boxes #6506

ahmad4633 commented Feb 2, 2022 •

edited by UltralyticsAssistant

Loading

github-actions bot left a comment

ahmad4633 commented Feb 15, 2022

tujh2 commented Feb 15, 2022 •

edited

Loading

RoudyES commented Feb 15, 2022

sctrueew commented Feb 15, 2022 •

edited

Loading

RoudyES commented Feb 15, 2022

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022

sctrueew commented Feb 15, 2022

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022 •

edited

Loading

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 16, 2022

sctrueew commented Feb 16, 2022

RoudyES commented Feb 16, 2022

sctrueew commented Feb 16, 2022

RoudyES commented Feb 16, 2022

GivanTsai commented Feb 18, 2022

GivanTsai commented Feb 18, 2022

github-actions bot commented Mar 21, 2022

Aun0124 commented Jan 11, 2023

tujh2 commented Jan 11, 2023

Aun0124 commented Jan 16, 2023

Adding polygon detection alongside bounding boxes #6506

Adding polygon detection alongside bounding boxes #6506

Conversation

ahmad4633 commented Feb 2, 2022 • edited by UltralyticsAssistant Loading

🛠️ PR Summary

🌟 Summary

📊 Key Changes

🎯 Purpose & Impact

github-actions bot left a comment

Choose a reason for hiding this comment

ahmad4633 commented Feb 15, 2022

tujh2 commented Feb 15, 2022 • edited Loading

RoudyES commented Feb 15, 2022

sctrueew commented Feb 15, 2022 • edited Loading

RoudyES commented Feb 15, 2022

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022

sctrueew commented Feb 15, 2022

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022 • edited Loading

sctrueew commented Feb 15, 2022 • edited Loading

tujh2 commented Feb 15, 2022 • edited Loading

sctrueew commented Feb 15, 2022 • edited Loading

sctrueew commented Feb 15, 2022

tujh2 commented Feb 15, 2022 • edited Loading

sctrueew commented Feb 16, 2022

sctrueew commented Feb 16, 2022

RoudyES commented Feb 16, 2022

sctrueew commented Feb 16, 2022

RoudyES commented Feb 16, 2022

GivanTsai commented Feb 18, 2022

GivanTsai commented Feb 18, 2022

github-actions bot commented Mar 21, 2022

Aun0124 commented Jan 11, 2023

tujh2 commented Jan 11, 2023

Aun0124 commented Jan 16, 2023

ahmad4633 commented Feb 2, 2022 •

edited by UltralyticsAssistant

Loading

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022 •

edited

Loading

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022 •

edited

Loading

tujh2 commented Feb 15, 2022 •

edited

Loading

sctrueew commented Feb 15, 2022 •

edited

Loading

tujh2 commented Feb 15, 2022 •

edited

Loading