Flowchart: The flow inside the module
Call-in on Imagine Marketplace and follow the Tutorial.
Important to remember here: if you have coding experience through the flowchart, use 'Jupyter', if not, use 'Deepaas' when you fill in the general information.
General configuration:
-
Deployment title: give name
-
Deployment description: add description
-
service to run: Jupyter, give pw
-
docker tag: CPU (can also use GPU or latest)
-
how do users know what to do, deepaas or jupyter? → jupyter for now, docker tag CPU
deepaas: Only activates the API to run the module, can be convenient if user has zero experience with coding and has all their data on nextcloud already Jupyter: accesses jupyter notebook from which deepaas can be ran with this command: nohup deep-start --deepaas &
Hardware configuration:
- Number of CPU: 1 for inference or 4/8 for training (max 10)
- Number of GPU: 0 (or 1 for training while combining the docker GPU tag) – not many available
- GPU model: all options are good
- RAM: 8000 MB default (max 40 GB)
- Disk memory: 10000 MB default (max 50 GB)
Data configuration advanced settings
- storage: /srv/.rclone/rclone.conf
- Rclone vendor: nextcloud
- storage URL: https://share.services.ai4os.eu/remote.php/webdav/
- RCLONE user and password:
- log in to https://share.services.ai4os.eu/index.php/login?redirect_url=/index.php/apps/dashboard/ using EGI and use your institutes IDP if possible
- in account go to profile > security > devices and sessions
Dataset configurations Pull in in data from zenodo/doi.
User Group error
- https://aai.egi.eu/registry/co_petitions/start/coef:550
- https://aai.egi.eu/registry/co_petitions/start/coef:546
You can train your own audio classifier with your custom dataset. For that you have to:
The first step to train you image classifier if to have the data correctly set up.
The model needs to be able to access the images. Please use a standard image format (like .png
or .jpg
).
So you either have to place your images in the ./data/images folder or if you have your data somewhere else you can use that location by setting the image_dir
parameter in the training args, for example, if you're data is on nextcloud, please change the parameter to: storage/name_folder/images
.
Another way is to copy the images
rclone copy rshare:/data/dataset_files /srv/phyto-plankton-classification/data/dataset_files
rclone copy rshare:/data/images /srv/phyto-plankton-classification/data/images
Next, you need add to the ./data/dataset_files directory the following files:
Mandatory files | Optional files |
---|---|
classes.txt , train.txt |
val.txt , test.txt , info.txt ,aphia_ids.txt |
The train.txt
, val.txt
and test.txt
files associate an image name (or relative path) to a label number (that has
to start at zero).
The classes.txt
file translates those label numbers to label names.
The aphia_ids.txt
file translates those the classes to their corresponding aphia_ids.
Finally the info.txt
allows you to provide information (like number of images in the database) about each class.
You can find examples of these files at ./data/demo-dataset_files.
If you don't want to create your own datasplit, this will be done automatically for you with a 80% train, 10% validation, and 10% test split.
Clarify the location of the images inside the yaml file file. If not, ./data/images will be taken. Any additional parameter can also be changed here such as the type of split for training/validation/testing, batch size, etc
You can change the config file directly as shown below, or you can change it when running the api.
images_directory:
value: "/srv/phyto-plankton-classification/data/images"
type: "str"
help: >
Base directory for images. If the path is relative, it will be appended to the package path.
After this, you can go to /srv/phyto-plankton-classification/planktonclas#
and run train_runfile.py
.
cd /srv/phyto-plankton-classification/planktonclas`
python train_runfile.py
The new model will be saved under phyto-plankton-classification/models
In stead of running everything within Jupyter, you can also run the following command:
nohup deep-start --deepaas &
Now you can go back to the deployments and run the API and change the paramters there. It is the same as method 2.3 for training, but more costumized.
Similar to 2.1.2: Running the training,clarify the location of the images inside the yaml file file. If not, ./data/images will be taken. Any additional parameter can also be changed here such as the type of split for training/validation/testing, batch size, etc
You can change the config file directly as shown below.
images_directory:
value: "/srv/phyto-plankton-classification/data/images"
type: "str"
help: >
Base directory for images. If the path is relative, it will be appended to the package path.
You can have more info on how to interact directly with the module (not through the DEEPaaS API) by examining the
./notebooks
folder:
- dataset exploration notebook: Visualize relevant statistics that will help you to modify the training args.
- Image transformation notebook: To conform a new dataset with the training set that was used
- Image transformation notebook: Notebook to perform image augmentation and expand the dataset.
- Model training notebook: Notebook to perform image augmentation and expand the dataset.
- computing predictions notebook: Test the classifier on a number of tasks: predict a single local image (or url), predict multiple images (or urls), merge the predictions of a multi-image single observation, etc.
- predictions statistics notebook:
Make and store the predictions of the
test.txt
file (if you provided one). Once you have done that you can visualize the statistics of the predictions like popular metrics (accuracy, recall, precision, f1-score), the confusion matrix, etc.
In stead of running everything within the notebooks, you can also run the following command:
nohup deep-start --deepaas &
Now you can go back to the deployments and run the API and change the paramters there. It is the same as method 2.3 for training, but more costumized.
Open deepaas through the API link in your deployment.
Look for the methods belonging to the planktonclas
module.
Look for the TRAIN
POST method. Click on 'Try it out', change whatever training args
you want and click 'Execute'. The training will be launched and you will be able to follow its status by executing the
TRAIN
GET method which will also give a history of all trainings previously executed.
You can follow the training monitoring (Tensorboard) by going back to the deployment page and selecting: monitor
Select the image or images you wanne predict.
You can have more info on how to interact directly with the module (not through the DEEPaaS API) by examining the
./notebooks
folder:
- computing predictions notebook: Test the classifier on a number of tasks: predict a single local image (or url), predict multiple images (or urls), merge the predictions of a multi-image single observation, etc.
- predictions statistics notebook:
Make and store the predictions of the
test.txt
file (if you provided one). Once you have done that you can visualize the statistics of the predictions like popular metrics (accuracy, recall, precision, f1-score), the confusion matrix, etc.
In stead of running everything within the notebooks, you can also run the following command:
nohup deep-start --deepaas &
Now you can go back to the deployments and run the API and change the parameters there when you look for the PREDICT
POST method. Click on 'Try it out', change whatever test args
you want and click 'Execute'. You can either supply a:
-
a
image
argument with a path pointing to an image. -
a
zip
argument with an URL pointing to zipped folder with images.
You can follow the Oscar Tutorial. But basically go to: https://inference.cloud.ai4eosc.eu/ui/#/login and sign in and look for the 'phyto-plankton-classifier'.
If you don't have an account yet then:
-
go to https://aai.egi.eu/ and login via the account you used for the platform (usually your institute)
-
go to top-right corner, you should see your name. Click on it and select "My EGI User Community Identity"
-
on the opened page, scroll down to "Role Attributes" and check if you find in the table a row with "vo.imagine-ai.eu". The "Status" must be "Active", also "Valid Through" should be in future.
-
IF you DON'T see "vo.imagine-ai.eu" listed, please, apply for the membership by clicking: https://aai.egi.eu/registry/co_petitions/start/coef:546