How to Train YOLOv7 On a Custom Dataset

In this video we walk through how to train YOLOv7 on your custom dataset.

1. What's New in YOLOv7
2. Exploring Roboflow Universe for example projects
4. Colab Tutorial on how to train YOLOv7

Resources:

[emptyingthecup]

A question on dataset images. If you're capturing images from a game, and the game screenshots are at 2560x1440, should you crop them to 640x640 before annotating them and building your final dataset for training? Or should you take the large dataset of 2560x1440 images, annotate them, and then feed them to the neural network for training? I noticed the images here are quite large, but when I download datasets from Roboflow, they're clearly cropped and/or resized. Thank you.

[AUDREYCARE]

how can I use this for the integrated webcam?

[alperenkuzhan]

How can i upload my yolov7 training to roboflow? The code at 13:14 doesn't work and the model does not appear in roboflow. I spent my entire 2 days just for this, please help.

[isnaayu_m]

Hi, would you tell me, what's the differents between yolov7_training and yolov7 ? I wanna training my custom dataset ? because I'm a little hesitant to use which model, thanks in advance

[muhdnajhan]

sir, how to put coordinate on the bounding box?

[harisumant]

Loved the video! How can I add an attention block to the architecture of YOLOv7 and train it on custom dataset?Are there any references for this task?

[akshayakki]

Sir, can you please tell me how to calculate the overall accuracy of the particular trained yolov7 model after testing an image in Google colab.

[chinthagovardhan]

I did with my custom dataset but when I detecting it not detecting in output what is the solution for that

[devfordevs]

how we convert/export this model to tflite model for using in edge device

[zombiefreak]

Hello, I have a question, is it possible to do this while using the cmd prompt only if i dont wanna use the google colab? Cuz i find it easier when im using the prompt itself

[moahaimen]

how to draw the loss and accuracy diagrams

[jozbrpw]

Hi, How can I continue training the yolov7 after fully completing the first time of training? Thanks

[mhmahfuz]

please give me the test command for testing dataset.

[jboxept]

heeey people does anybody know how to store the train results in a csv file ?

[henkiespenkie]

does the dataset needs to be 416 by 416?

[ubuntulinux]

I am getting this kind of error what could be the reason
assert mlc < nc, 'Label class %g exceeds nc=%g in %s. Possible class labels are 0-%g' % (mlc, nc, opt.data, nc - 1)
AssertionError: Label class 15 exceeds nc=5 in data/custom.yaml. Possible class labels are 0-4

[ytkocrew]

he said nothing about remake data/"yourname".yaml and cfg/"yourname".yaml files for using yolov7 on custom data.
zero information of that. but without this you'll can't do a training.

[abrarluvrabit]

Can we also do active learning in our coding like out of 20 testing images my system detect 19 correct images but only 1 incorrect detection so how can i tell my system about this 1 incorrect image detection so next time it will not repeat the same mistake again. and how can i do that without training the entire model again?

[zeinabmoradi]

please update the zero-shot tracker code in order to use it with this detector yolov7

[rohanpokale]

I am getting this error can you pls help

Epoch gpu_mem box obj cls total labels img_size
0% 0/15 [00:13<?, ?it/s]
Traceback (most recent call last):
File "train.py", line 616, in <module>
train(hyp, opt, device, tb_writer)
File "train.py", line 363, in train
loss, loss_items = compute_loss_ota(pred, targets.to(device), imgs) # loss scaled by batch_size
File "/content/yolov7/utils/loss.py", line 585, in __call__
bs, as_, gjs, gis, targets, anchors = self.build_targets(p, targets, imgs)
File "/content/yolov7/utils/loss.py", line 759, in build_targets
from_which_layer =
RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)