Skip to main content
Version: Next

VC Release Note

v1.5.3​

May. 30, 2024

Improvements Remove unused functions

  • Remove check_memory

Bug Fixes Update for inference_summary.yaml

  • Change key format of probability

Compatibility: ALO v2.5.1 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.5.2​

May. 30, 2024

Improvements Rename content's name

  • Change content's name from Computer Vision AI to Vision Classification

Compatibility: ALO v2.3.4 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.5.1​

May. 10, 2024

Improvements Change output image size

  • Instead of resized image, save original image into output folder for showing the image at edge application

Bug Fixes Fix saving image color bug

  • Change image saving function with RGB

Compatibility: ALO v2.3.3 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.5.0​

Apr. 17, 2024

Improvements Improve inference runtime

  • Achieving Approximate 1-Second Inference Time in Splunk Edge Hub(SEH) Execution

Compatibility: ALO v2.4.0 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.4.0​

Mar. 29, 2024

Bug Fixes Change image path using relative path

  • When image path is written by '/nas001/.../image.png' closed system can't access to such format

Fix data type error in ui_args

  • Add function to change ui_args data type

Compatibility: ALO v2.3.1 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.3.1​

Mar. 14, 2024

New Features Checking memory and runtime

  • Add check memory and runtime function to train and inference asset

Improvements Update for Mellerikat requirements

  • Support ALO2.3

Bug Fixes Available on solution registration

  • Fix ui_args_detail's argument naming error

Compatibility: ALO v2.3 Tested on c6i.2xlarge - CPU: 6 / MEM: 14G, p3.2xlarge - CPU: 6 / MEM: 55G / GPU: 1(16G) / CUDA 11

v1.3.0​

Feb. 29, 2024

New Features

Explainable AI(XAI)

  • Added functionality to assist both end-users (Edge) and analysts (Conductor) in understanding why a certain classification occurred based on specific areas within the image.  

Improvements Update for Mellerikat requirements

  • Move extra output files
  • Erase label_names, y_column, path_column user parameters from experimental_plan.yaml
  • Use ALO's logging api

Bug Fixes

Clean code and Fix data error

  • Erase unnecessary functions
  • Fix channel duplicating error

Compatibility: ALO v2.2.1

v1.2.1​

Jan. 19, 2024

Bug Fixes

  • Fix data type of score as float
  • Fix error saving score in inference_summary.yaml

v1.2.0​

Jan. 16, 2024

Bug Fixes Fix importance error

  • Fix probability/prediction prefix setting error  during inference pipeline

Improvements Support checkpoint function

  • During training, to save and load best model only

Add uncertainty column

  • Add uncertainty column into output file: prediction.csv. It makes that user can recognize which data needs relabeling.

Update for Mellerikat requirements

  • Save inference_summary.yaml for both batch and single data

Compatibility: ALO v2.1

v1.1.1​

Nov. 10, 2023

Bug Fixes Fix inference error

Improvements Number of argument for training step reduced

  • Updates for user convenience

v1.1.0​

Nov. 06, 2023

Bug Fixes Fix training error

Compatibility: ALO v2.0

v1.0.2​

Nov. 01, 2023

Improvements Enhancement augmentation method

  • Excluding multiple augmentation methods is available
  • Changing magnitude of augmentation is available
  • Changing number of how many times image will be augmented is available

v1.0.1​

Oct. 27, 2023

New Features Updates for Mellerikat requirements

  • Support .tflite model
  • Enable sample notebook

Improvements Improve output dictionary's format in ic asset

Compatibility: ALO v1.1

v1.0.0​

Oct. 23, 2023

New Features Add image classification models

  • basic model is available: mobilnet V1
  • High resolution model is available: customized mobilenet V3

random augmentation is available

  • random augmentation makes model can recognize pictures taken in various condition

Improvements Efficient training/inferencing pipeline

  • prefetch on CPU and GPU is available

Compatibility: ALO v1.0