visualone

Visual One

Visual One's few shot learning framework allows you to easily train models for complex computer vision tasks using only a few samples in two lines of code.

To see some examples, visit the examples page of our website.

Quickstart

• Install the package: pip install visualone

• Submit your email here to receive your public and private keys via email.

• Train a model by providing some positive samples and some negative samples:

from visualone import vedx

client = vedx.client(public_key, private_key)

model = client.train(positive_samples, negative_samples)

• Apply the trained model to a new image to do prediction:

client.predict(model['task_id'], image_file)

Description of inputs & outputs

Training

model = client.train(positive_samples, negative_samples)

positive_samples and negative_samples must be either:

• str: The path to a directory containing image files representing all of the positive/negative samples.
  Or
  
• list[str]: The name of the individual files corresponding to the positive/negative samples.

Returns a dict with the following keys:
task_id: A unique task id generated for this task. You need to pass this to client.predict to do prediction.
n_positive_samples: The number of positive samples used for training the model.
n_negative_samples: The number of negative samples used for training the model.
success: 1 if the training was done successfully. 0 if an error occured.
message: A brief message describing the error if an error occurs.

Prediction

client.predict(task_id, image_file)

task_id: The unique task id corresponding to the trained model returned by vedx.train
image_file: The path to the file for which you want to do inference.

Returns a dict with the following keys:
task_id: The task id corresponding to the model.
prediction: A boolean representing the model prediction. True means the model predicted a positive label for the given image. False means the model predicted a negative label for the given image.
confidence: A number between 0 and 100 representing the confidence of the model for this prediction.
latency: The time it took to do the prediction in millisecond.

Report Bugs/Issues & Feature Request

Please, help us improve our product by reporting any bugs or issues while using visualone. Also, feel free to let us know any feature requests.