Hello everyone ,
I want to ask you about maximum size of training dataset ? is it 4GB? is there anyway to increase the size in Edge Impulse ?
Hi @Bayan_khalid,
I do not believe there is a limit to the dataset size. However, you are limited to 4 GB of RAM when doing DSP preprocessing, 8 GB of RAM during training, and 20 min of processing (DSP or training) time. Using a large dataset will likely make you hit one of those limits.
As for increasing those limits in the free version:
- If it is for a sustainable cause, I can up your limit if you explain the details of your project here.
- If it is for a commercial project, I invite you to contact our sales team here: Pricing
- If it is for testing purposes, you can use the tips outlined on this page: Lower compute time - Edge Impulse Documentation
- Also note, if you are willing to share your project publicly, you can put a nice description on the your project’s description (https://www.edgeimpulse.com/blog/showcase-your-public-projects-with-markdown-readmes) with as much context as you can (goal of this project, etc.), and I can review it to increase your time limit.
Yes, clear
Does this include the education purposes(Senior Graduation Project).?
Best Regards
Hi @shawn_edgeimpulse,
I would like to upgrade my limit for my free version. I am a student working on a final year project in India.
We are working on a waste segregation robot that separates plastic bottles from other recyclable waste.This plastic bottles will be further recycled for the better management of plastic waste. We are using The FOMO (Faster Objects, More Objects) MobileNetV2 0.35 for object detection.
Currently our dataset contains 1266 images in the dataset.
We tried to train using the following parameters:
- Number of training cycles: 70
- Learning rate: 0.008
- Batch size: 32
It gave us an error, Processing time was 26 mins est. which was greater than the 20 min limit.
I want to ask if this parameter of number of training cycles is good for detecting objects in my project.
I want to ask if this parameter of number of training cycles is good for detecting objects in my project.
Number of epochs (training cycles) can vary widely from project to project based on the complexity of model, amount of data, and complexity of that data. Your best bet is some trial and error. Start with a high number (like 100) and see how well the model performs. Does it converge to an accuracy number prior to reaching that 100th cycle? Or is the loss/accuracy still changing a lot?
