Comments (6)
I realized that there are several parameters to adjust, such as the outputs the firmware and maybe it is interesting to say, (at least for me because all the models I trained gave a problem because the kmodel output is always greater than 3mb) anyway, I think it is interesting say that:
Step 1 - Use the minimum firmware with the latest ide with v4 support to use in Maixpy or without ide directly in the terminal (I couldn't see the output in the terminal but maybe it is from the script, I will study it yet).
Step 2 - The script for v4 that is here in your example scripts works
Step 3 - But it only works for sd card loading (all models I trained gave more than 3mb)
Step 4 - Between tests always remember to press the reset button
Step 5 - When training, it is easier to use Roboflow.com to resize the images and store them.
Great job, I used transferlearning too and I'm delivering a great job to my college in Brazil
from axelerate.
Eu e meus colegas de equipe trabalhamos em dois projetos usando MobileNet1_0 em 2 modelos, um era classificação e o outro era detecção de objetos. Eram ajustes apertados e tivemos que usar o
maixpy_v0.6.2_41_g02d12688e_minimum_with_kmodel_v4_support
firmware customizado para fazê-lo funcionar.A classificação de objetos foi apenas um teste básico, provavelmente mais de 150 imagens, mas a classificação estávamos usando quase 2.000 imagens e funcionou bem.
Salve irmão
- HAHAHA adorei o Salve irmão
I also used this with the ide, it worked well here, I used a pre-trained network to improve the performance with 430 images, maybe I should increase the bank or not because I will already deliver this work.
from axelerate.
Use the minimum firmware with the latest ide with v4 support to use in Maixpy or without ide directly in the terminal (I couldn't see the output in the terminal but maybe it is from the script, I will study it yet).
Yes. I normally build firmware myself, with kmodelv4 support, IDE and optionally ulab/video. It is ~1.6 Mb in size.
The script for v4 that is here in your example scripts works
Yes. aXeleRate uses nncase v0.2.0-beta4, which outputs kmodelv4
But it only works for sd card loading (all models I trained gave more than 3mb)
If you plan to use Micropython firmware, you should use MobileNet2_5, MobileNet5_0,, MobileNet7_5 or Tiny Yolo backends. You probably have used MobileNet1_0, which is too large to fit in memory IF Micropython is used. MobileNet1_0 can be used if you're writing C code for K210.
Between tests always remember to press the reset button
Yes.
When training, it is easier to use Roboflow.com to resize the images and store them.
Not sure why. aXeleRate automatically resizes and even augments the images.
from axelerate.
Anyways, it should work with latest version of Micropython firmware with kmodel v4 support.
The memory needed to run the model depends on backend chosen.
Does it answer your question?
from axelerate.
Perfectly.
I will change the backend, Thanks
from axelerate.
Me and my teammates worked on two projects using MobileNet1_0 on 2 models, one was classification and the other was object detection. These were tight fits, and we had to use the maixpy_v0.6.2_41_g02d12688e_minimum_with_kmodel_v4_support
custom firmware to make it work.
The object classification one was just a basic test, probably 150+ images, but the classification one we were using almost 2000 images, and it worked fine.
Salve irmão
from axelerate.
Related Issues (20)
- 是否支持把没有目标的空背景和标签输入训练? HOT 2
- Does training with negative samples(pictures without objects of interest) increase accuracy? HOT 6
- Can the aXeleRate support the format? HOT 1
- k210 segnet HOT 1
- json.decoder.JSONDecodeError: Expecting ',' delimiter: line 35 column 2 (char 1325) HOT 5
- Cannot convert tf to onnx in object detection HOT 2
- Training person detector with pascal_20_detection dataset error HOT 9
- tensorflow 2.5 error! HOT 1
- Support for custom input_size no work in kmodel HOT 4
- [unstable branch] validation frequency adding to config.json HOT 1
- [unstable branch] yolo training quantize or not can be switched by config.json HOT 1
- [unstable branch] yolo k210 converter failed while nncase compile and YOLOv3 convert incorrectly HOT 1
- Threads Error HOT 1
- The loss is not converge when training detector on VOC 2012 HOT 2
- When class names more than one, the mAP is false? HOT 5
- How calculate anchors by kmeans? HOT 2
- Multi Object Detect Is Wrong! HOT 2
- Loading kmodel on MaixPy crashing device HOT 25
- ValueError: Invalid value for argument filters. Expected a strictly positive value. Received filters=0. HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from axelerate.