CustomVisionAI 5 Day Training for 90 Images

Hello,

It is indeed unusual for a Custom Vision AI model to take more than 5 days to train on just 90 images, especially since the "Free Tier" usually processes small datasets quickly.

Here are a few steps to resolve this:

Check Training Type: If you selected "Advanced Training," it can take a long time. For 90 images, try using "Quick Training" instead.

Stuck in Queue: I recommend canceling the current training session and starting a new one. Often, a fresh start clears the process within minutes.

Check Resource Status: Ensure your Azure resource hasn't hit any student-plan limits.

Hope this helps!

Post Your Answer: जवाब लिखने के बाद, उसके ठीक नीचे जो नीले रंग का बटन है "Post your answer",

Hey LucienSP1,

Thank you for reaching out on the Microsoft Q&A.

It sounds like you're running into a frustrating situation with your Custom Vision model training taking longer than expected, especially given your dataset of 90 images. Here are a few things you might want to consider troubleshooting this issue:

Review Training Data: Ensure your images are compliant with privacy standards. If any of your images contain faces, it may cause warnings and impact the training efficiency. You can navigate to your dataset in the Azure portal to check for flagged images and modify your dataset as necessary.

Model Complexity and Resources: Depending on the model complexity and the resources allocated for training, the duration can vary. Ensure that you have selected optimal compute resources for your training. If you are on the free tier, there might be limitations in terms of resource allocation.

Training Parameters: Double-check the model parameters you are using for training. Sometimes, tweaking these can optimize the training duration. Make sure you are also aware of any quotas on the free tier – they may hinder your training efficiency.

Try Retraining: If the training is stuck or taking unusually long, you could try stopping the current training session and starting from scratch, ensuring the training data is set up correctly.

Cost Monitoring: If you're approaching any limits on your current tier, consider upgrading to a Standard Tier (S0) to avoid potential bottlenecks.

I Hope this helps. Do let me know if you have any further queries.

References:

• Troubleshoot Image Vision Fine-Tuning Issues
• Set up AutoML to train computer vision models with Python
• Custom Vision pricing

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thankyou!