YoloModel - Custom Model

[alijundi]

Hi @kkarahainko @keesschollaart81

I added my custom model and exported my .onnx from colab I ran the project after making my custom model.

I get System.IndexOutOfRangeException: 'Index was outside the bounds of the array.'
in function ParseDetect. I am not sure what to change in my custom model class to avoid this!

By the way. The code works well on the models and test image provided.

[kkarahainko]

@alijundi please provide a screenshot from Netron (like this https://i.imgur.com/2X2H2Co.png) and your custom model class (code) with parameters.

[alijundi]

@kkarahainko

Thank you for your prompt answer. Below is the code:

`using System.Collections.Generic;
using Yolov5Net.Scorer.Models.Abstract;

namespace Yolov5Net.Scorer.Models
{
public class YoloCocoP7Model : YoloModel
{
public override int Width { get; set; } = 416;
public override int Height { get; set; } = 416;
public override int Depth { get; set; } = 3;

    public override int Dimensions { get; set; } = 85;

    public override float[] Strides { get; set; } = new float[] { 8, 16, 32, 64 };

    public override float[][][] Anchors { get; set; } = new float[][][]
    {
        new float[][] { new float[] { 019, 027 }, new float[] { 044, 040 }, new float[] { 038, 094 } },
        new float[][] { new float[] { 096, 068 }, new float[] { 086, 152 }, new float[] { 180, 137 } },
        new float[][] { new float[] { 140, 301 }, new float[] { 303, 264 }, new float[] { 238, 542 } },
        new float[][] { new float[] { 436, 615 }, new float[] { 739, 380 }, new float[] { 925, 792 } }
    };

    public override int[] Shapes { get; set; } = new int[] { 20, 40, 80, 160 };

    public override float Confidence { get; set; } = 0.20f;
    public override float MulConfidence { get; set; } = 0.25f;
    public override float Overlap { get; set; } = 0.45f;

    public override string[] Outputs { get; set; } = new[] { "output" };

    public override List<YoloLabel> Labels { get; set; } = new List<YoloLabel>()
    {
        new YoloLabel { Id = 15, Name = "FEEDER_MCCB_3P_100A" },
        new YoloLabel { Id = 16, Name = "FEEDER_MCCB_3P_40A" },
        new YoloLabel { Id = 17, Name = "INCOMER_MCCB_3P_160A" },
    };

    public override bool UseDetect { get; set; } = true;

    public YoloCocoP7Model()
    {

    }
}

}
`
.onnx file download link: https://drive.google.com/file/d/1-4TR22metrnzV9JXHuzstrKLHf5L7v7M/view?usp=sharing
Netron image file: https://drive.google.com/file/d/1PzZvqlOaYoZjxoKSHOmy-S-YWEiffi45/view?usp=sharing

[alijundi]

using System.Collections.Generic;
using Yolov5Net.Scorer.Models.Abstract;

namespace Yolov5Net.Scorer.Models
{
    public class YoloCocoP7Model : YoloModel
    {
        public override int Width { get; set; } = 416;
        public override int Height { get; set; } = 416;
        public override int Depth { get; set; } = 3;

        public override int Dimensions { get; set; } = 85;

        public override float[] Strides { get; set; } = new float[] { 8, 16, 32, 64 };

        public override float[][][] Anchors { get; set; } = new float[][][]
        {
            new float[][] { new float[] { 019, 027 }, new float[] { 044, 040 }, new float[] { 038, 094 } },
            new float[][] { new float[] { 096, 068 }, new float[] { 086, 152 }, new float[] { 180, 137 } },
            new float[][] { new float[] { 140, 301 }, new float[] { 303, 264 }, new float[] { 238, 542 } },
            new float[][] { new float[] { 436, 615 }, new float[] { 739, 380 }, new float[] { 925, 792 } }
        };

        public override int[] Shapes { get; set; } = new int[] { 20, 40, 80, 160 };

        public override float Confidence { get; set; } = 0.20f;
        public override float MulConfidence { get; set; } = 0.25f;
        public override float Overlap { get; set; } = 0.45f;

        public override string[] Outputs { get; set; } = new[] { "output" };

        public override List<YoloLabel> Labels { get; set; } = new List<YoloLabel>()
        {
            new YoloLabel { Id = 15, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 16, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 17, Name = "INCOMER_MCCB_3P_160A" },
        };

        public override bool UseDetect { get; set; } = true;

        public YoloCocoP7Model()
        {

        }
    }
}

[kkarahainko]

@alijundi

1. you have no Detect layer exported (detect layer looks like this https://i.imgur.com/FgjTDTQ.png)
2. wrong dimensions, should be 36 instead of 85, wrong shapes, should be { 13, 26, 52 } instead of { 20, 40, 80 };
3. looks like your model trained on yolov5 P5 version but you use P6 code as a reference, use YoloCocoP5Model example as a reference;
4. you have 31 labels but I see only 3 labels in your model class, all labels should be described;

my recommendation - clone and use the latest ultralytics/yolov5 version in collab, it exports Detect layer by default, I'm not sure you've exported your model right.

so, your model class should look like this:

public class MyCustomModel : YoloModel
{
	public override int Width { get; set; } = 416;
	public override int Height { get; set; } = 416;
	public override int Depth { get; set; } = 3;

	public override int Dimensions { get; set; } = 36;

	public override float[] Strides { get; set; } = new float[] { 8, 16, 32 };

	public override float[][][] Anchors { get; set; } = new float[][][]
	{
		new float[][] { new float[] { 010, 13 }, new float[] { 016, 030 }, new float[] { 033, 023 } },
		new float[][] { new float[] { 030, 61 }, new float[] { 062, 045 }, new float[] { 059, 119 } },
		new float[][] { new float[] { 116, 90 }, new float[] { 156, 198 }, new float[] { 373, 326 } }
	};

	public override int[] Shapes { get; set; } = new int[] { 13, 26, 52 };

	public override float Confidence { get; set; } = 0.20f;
	public override float MulConfidence { get; set; } = 0.25f;
	public override float Overlap { get; set; } = 0.45f;

	public override string[] Outputs { get; set; } = new[] { "output", "465", "485" };

	public override List<YoloLabel> Labels { get; set; } = new List<YoloLabel>()
	{
		// all 31 lables described here
	};

	public override bool UseDetect { get; set; } = false;

	public YoloCocoP5Model()
	{

	}
}

[alijundi]

@kkarahainko Thank you for your feedback:

I updated the code to the below:

[alijundi]

using System.Collections.Generic;
using Yolov5Net.Scorer.Models.Abstract;

namespace Yolov5Net.Scorer.Models
{
    public class YoloCocoP7Model : YoloModel
    {
        public override int Width { get; set; } = 416;
        public override int Height { get; set; } = 416;
        public override int Depth { get; set; } = 3;

        public override int Dimensions { get; set; } = 36;

        public override float[] Strides { get; set; } = new float[] { 8, 16, 32 };

        public override float[][][] Anchors { get; set; } = new float[][][]
        {
            new float[][] { new float[] { 010, 13 }, new float[] { 016, 030 }, new float[] { 033, 023 } },
            new float[][] { new float[] { 030, 61 }, new float[] { 062, 045 }, new float[] { 059, 119 } },
            new float[][] { new float[] { 116, 90 }, new float[] { 156, 198 }, new float[] { 373, 326 } }
        };

        public override int[] Shapes { get; set; } = new int[] { 13, 26, 52 };

        public override float Confidence { get; set; } = 0.20f;
        public override float MulConfidence { get; set; } = 0.25f;
        public override float Overlap { get; set; } = 0.45f;

        public override string[] Outputs { get; set; } = new[] { "output", "465", "485" };

        public override List<YoloLabel> Labels { get; set; } = new List<YoloLabel>()
        {
            new YoloLabel { Id = 0, Name = "DOG" },
            new YoloLabel { Id = 1, Name = "DOG" },
            new YoloLabel { Id = 2, Name = "DOG" },
            new YoloLabel { Id = 3, Name = "DOG" },
            new YoloLabel { Id = 4, Name = "DOG" },
            new YoloLabel { Id = 5, Name = "DOG" },
            new YoloLabel { Id = 6, Name = "DOG" },
            new YoloLabel { Id = 7, Name = "DOG" },
            new YoloLabel { Id = 8, Name = "DOG" },
            new YoloLabel { Id = 9, Name = "DOG" },
            new YoloLabel { Id = 10, Name = "DOG" },
            new YoloLabel { Id = 11, Name = "DOG" },
            new YoloLabel { Id = 12, Name = "DOG" },
            new YoloLabel { Id = 13, Name = "DOG" },
            new YoloLabel { Id = 14, Name = "DOG" },
            new YoloLabel { Id = 15, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 16, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 17, Name = "INCOMER_MCCB_3P_160A" },
            new YoloLabel { Id = 18, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 19, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 20, Name = "INCOMER_MCCB_3P_160A" },
            new YoloLabel { Id = 21, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 22, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 23, Name = "INCOMER_MCCB_3P_160A" },
            new YoloLabel { Id = 24, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 25, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 26, Name = "INCOMER_MCCB_3P_160A" },
            new YoloLabel { Id = 27, Name = "FEEDER_MCCB_3P_100A" },
            new YoloLabel { Id = 28, Name = "FEEDER_MCCB_3P_40A" },
            new YoloLabel { Id = 29, Name = "INCOMER_MCCB_3P_160A" },
            new YoloLabel { Id = 30, Name = "FEEDER_MCCB_3P_100A" }
        };

        public override bool UseDetect { get; set; } = false;

        public YoloCocoP7Model()
        {

        }
    }
}

[alijundi]

[alijundi]

@kkarahainko I am getting the above exception now. Am I doings something wrong here?

[alijundi]

I used the below code in colab to export .pt trained model to .onnx

!python /content/yolov5/models/export.py --weights /content/yolov5/runs/train/yolov5s_results/weights/best.pt --img 416 --batch 1 # export

Any thoughts?

[kkarahainko]

@alijundi pls, change Shapes order "public override int[] Shapes { get; set; } = new int[] { 52, 26, 13 };"

[kkarahainko]

@alijundi can you provide a test image?

[alijundi]

@kkarahainko lemme try it now...

[alijundi]

@kkarahainko result is being outputed with no predictions.. INPUT == OUTPUT

[alijundi]

[alijundi]

above is test input

[kkarahainko]

@alijundi is yolov5/detect.py detects something? did you try?

[alijundi]

[alijundi]

The output is from Roboflow-Custom-YOLOv5 that is what I used to train my model

[kkarahainko]

@alijundi I mean did you try detect.py on the test image you've sent to me?

[kkarahainko]

@alijundi I suppose it doesn't detect too?

[alijundi]

@kkarahainko I am sorry I took sometime to respond to you. I was actually on the way home.

and YES! I tried all images that make up the model with no success. It is not detecting anything!

Do you know why?

[kkarahainko]

@alijundi it's all about pixels, the test image is b/w no labels etc, training image is colored with labels, you should train on the similar images you will use further (to predict), or not enough epochs, low mAP etc.