Best Number of epochs?

[yazeedhasan97]

Hi,

I am training my model on a data set with 850 train image and 250 val image

the thing is, I am running the training for 30 epochs and 12 batches (that what my device can take, and take around 3 hours to finish), but still, the mAP is very low and the model poorly detects the objects I am training on

I am gathering and labeling more data for the training process in the mean time. However, I have 3 questions?

Are the number of epochs related to the amount of my training data(more data/more epochs OR less data/more epochs)?
Do low batches affect the training mAP?
What is the best number of epochs for such dataset (850 train image and 250 val)

Please find the result of my training

  0/29        0G   0.09966   0.01311         0    0.1128         1       480         0         0 2.492e-06 5.397e-07   0.08867  0.009083         0
  1/29        0G   0.08995   0.01455         0    0.1045         1       352         0         0  0.001459 0.0001784   0.08163  0.009477         0
  2/29        0G   0.08862    0.0123         0    0.1009         1       480         0         0 0.0005077 7.246e-05   0.09102   0.01327         0
  3/29        0G   0.08621   0.01414         0    0.1004         1       224         0         0   0.01481  0.002737   0.07515   0.01159         0
  4/29        0G   0.08282   0.01361         0   0.09644         1       416  0.003232    0.4158  0.004293   0.00105   0.08466    0.1115         0
  5/29        0G   0.08837   0.01405         0    0.1024         1       480  0.008542     0.299  0.004305 0.0006745   0.09466   0.02414         0
  6/29        0G   0.08527   0.01391         0   0.09918         1       192         0         0  0.002861 0.0005437   0.08161   0.01095         0
  7/29        0G   0.07894   0.01498         0   0.09392         1       320   0.09511   0.07904   0.03146  0.006696   0.06468   0.01462         0
  8/29        0G   0.08419   0.01331         0    0.0975         1       288         0         0   0.04485   0.01217   0.07224   0.01132         0
  9/29        0G   0.08268   0.01297         0   0.09566         1       352         0         0   0.01232  0.001863   0.07967   0.01109         0
 10/29        0G   0.08223   0.01454         0   0.09677         1       416         0         0  0.004548 0.0007495   0.07592   0.01198         0
 11/29        0G   0.07916   0.01374         0    0.0929         1       224   0.05356    0.4296   0.09908   0.01612   0.06536   0.01399         0
 12/29        0G   0.07961   0.01356         0   0.09317         1       448   0.06935     0.512    0.1334   0.02706   0.06485   0.01415         0
 13/29        0G    0.0763   0.01324         0   0.08954         1       480   0.09109   0.03093   0.02091  0.002706   0.06997   0.01173         0
 14/29        0G   0.07497   0.01359         0   0.08856         1       256         0         0   0.02358  0.003817   0.06593    0.0119         0
 15/29        0G   0.07631    0.0125         0   0.08882         1       224         0         0   0.02196   0.00485   0.07033   0.01195         0
 16/29        0G   0.07737   0.01369         0   0.09106         1       192    0.0278   0.01031    0.0353  0.006509   0.06075   0.01252         0
 17/29        0G   0.07206   0.01327         0   0.08533         1       320   0.07842    0.2027   0.05105   0.01499   0.05978     0.012         0
 18/29        0G   0.07173   0.01361         0   0.08534         1       160    0.1081   0.01031   0.06596   0.01392   0.05635   0.01201         0
 19/29        0G   0.07049   0.01415         0   0.08464         1       480    0.1251    0.4983    0.1363   0.03508    0.0612   0.01218         0
 20/29        0G   0.06933   0.01355         0   0.08288         1       320   0.02504   0.01718   0.01769  0.003571   0.05766    0.0121         0
 21/29        0G   0.07087    0.0139         0   0.08477         1       192    0.1205   0.08591    0.0683   0.01458   0.06045   0.01111         0
 22/29        0G   0.06656   0.01429         0   0.08085         1       352   0.09393     0.512    0.1206    0.0279   0.05108   0.01217         0
 23/29        0G   0.06508   0.01382         0    0.0789         1       288     0.185     0.071   0.07345   0.01645   0.05313   0.01193         0
 24/29        0G   0.06433   0.01284         0   0.07717         1       320    0.1336    0.4261    0.1446   0.02885   0.05331   0.01156         0
 25/29        0G   0.06212   0.01339         0   0.07552         1       384    0.1365    0.4948    0.1529   0.03423   0.05144   0.01112         0
 26/29        0G   0.06044   0.01307         0   0.07351         1       192    0.1658     0.488    0.1893   0.05188   0.04968   0.01078         0
 27/29        0G   0.06347   0.01157         0   0.07504         1       288    0.1123    0.5189    0.1937   0.03736   0.05165    0.0116         0
 28/29        0G    0.0614   0.01307         0   0.07447         1       320    0.2084     0.299    0.1484    0.0376    0.0501   0.01064         0
 29/29        0G   0.05834   0.01317         0    0.0715         1       256    0.1034    0.4124   0.08006   0.02031   0.04898   0.01197         0

[github-actions]

Hello @yazeedhasan97, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

Get yourself a GPU, train to 300 epochs. We offer a wealth of free GPU environments, not sure why you would opt to train on CPU...

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab Notebook with free GPU:
• Kaggle Notebook with free GPU: https://www.kaggle.com/ultralytics/yolov5
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Docker Image https://hub.docker.com/r/ultralytics/yolov5. See Docker Quickstart Guide

[github-actions]

This issue 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.