Introducing Aliquis®

Aliquis® presents a rich set of features for prototyping and deploying real world applications for real time complex scene understanding and industrial manufacturing processing. With Aliquis® you can process 8bit/16bit/float multi-spectra (eg. RGB, RGB-IR, RGB-Xray, etc) volumetric data (3D), Z-stacked data (2.5D), planar data (2D aka image), and linead data (1D) both static (single time shot) and dynamic (multiple time shots) collected from a single point of view or multiple camera.

Aliquis® is in fact able to understand a real scene just like our own eyes and, consequently, make decisions based on what it has “seen”. It can therefore be applied to any decision-making activity based on vision. Aliquis® is built only on OpenCV and Python. We have developed Aliquis® with some simple concepts - patches, stages and pipelines - combined with a simple declarative language. These choices mean that even newbie users can quickly create prototypes and deploy industrial-level applications in the field of machine vision. Aliquis® supports Linux natively and Windows through virtual machines. Latest releases are shipped with Google TensorFlow™ throught Keras as preferred Deep Learning engine, while Berkeley Caffe is mantained for legacy applications.

Aliquis® in Capped Mode

Aliquis® 2.x.x in Capped Mode is a publicly available version of Aliquis®. This software contains all the most important building blocks of Aliquis® and can be used free-of-charge for both commercial and industrial applications. The Capped Mode is slowed down of about 10x thus it enables only near real-time applications.

The next major release 3.x.x is planned for Q2 2020.

Following documentation refers to Aliquis® in Capped Mode.

Please get in touch with us if you need further support for industrial applications.

Changelog of 2.x.x branch

## 2.3.2 [2019-10-03] - Added fine-tuning to aliquispl_keras - Added loss function infogain_categorical_crossentropy - Enhanced ximage_geom_3dstats - Multiple logdir creation in aliquispl_keras - Added flag -Z to aliquispl_run (zero image for ximages) - Bug fix in aliquispl_dset (info empty set)

## 2.3.1 [2019-09-13] - Added “disabled” flag to stages definition (bypass current stage)

## 2.3.0 [2019-09-06] - New HDF5 dataset management host aliquispl_dset - New Keras models training host aliquispl_keras - Added parallel pool to RANDOM_TRANSFORM stage - Minor bug fix in core stage initialization

## 2.2.4 [2019-07-10] - Enhanced decoration with aliquispl_run - New host: ximage_geom_2dstats

## 2.2.3 [2019-06-18] - New SHAPES_ON_ROOT stage semantics - Added loading of flattened shapes from ximage (ximage_meta: SHAPES_FLATTEN) - Added color_map to pipelines - Added Aliquis version info to aliquispl_run -v - Added -H flag (–hide-bbox) and -R (–force-rgb) flags to aliquispl_run - Bug fix in arithmetic-logic stage

## 2.2.2 [2019-05-27] - Fixed shapes forwarding in the STACK_CHANNELS stage - Added SEG-Y files support - Fixed issue

## 2.2.0 [2019-05-03] - Added heatmap_multiplier and heatmap_multiplier_stride in NN stage - Modified padding border type from REFLECT_101 to CONSTANT (black) in tools/dataset_creation/create_dataset - moved parameter remap_255_to_0 to conf, folder refactoring - Aligned tools/ximage/classes.xml to classes colors used in debugdraw

## 2.1.4 [2019-04-09] - XImage submodule sync

## 2.1.3 [2019-04-08] - Modified tool for model setup: from 400k to 800k train iterations, from 100 to 250 iteration test step, brightness data augmentation set to default (together with pca) - Updated tools/registration/ added check on input images (must be bitmask) - Updated tools/registration/ (improved debug results) - Removed debug print - Added support for multipage images - Dataset creation: added conf file example - Added data augmentation color shift flag to tools/caffe_train/

## 2.1.2 [2019-03-25] - Added aliquispl_benchmark host

## 2.1.1 [2019-03-22] - Added variable ALIQUIS_HOME for hosts

## 2.1.0 [2019-03-22] - Updated README - Modified tools/model_test/ - Added scale parameter to NN with Keras - Added script for evaluate statistics parameter of images - Added support for fully convolutional NN with Keras - Added Keras support - Added script for color gamma/scale correction of a folder of images - Added custom stages for RGB color transformation (scale and gamma) - Moved ximage scripts from tools/dataset_creation/voldemort/ximage_classes_batch to tools/ximage/ - Added/modified scripts for 2-classes problem FROC - Bug fix in fixed window stage - Modified tools/model_test/, added tools/model_test/

## 2.0.3 [2019-02-13] - Updated display messages for current year

## 2.0.2 [2019-02-13] - Added ignore_ximage_meta flag to SOURCE stage - Added capped mode - Path fix in CMake configuration script - added remap_255_to_0 parameter - Added new custom stage and refactoring - Modified aliquis.proto for proto3 compatibility - Modified stage heatmap: now the output is a cubic heatmap, like neural network stage heatmap - Bug fix in SGLock management - Bug fix in FILL_HOLES stage - Bug fix in padding stage

## 2.0.1 [2018-11-19] - Fixed conda/path issues - Added script for caffe convolution kernels visualization - Bug fix in default interpolation parameter - Added script for finding empty images filtering parameters (stats_filter stage) - Added Conda support - Modified tools/crf/ - Modified tools/nir/ added choice of y-axis range - CRF: update custom stage and added material - Bug fix in aliquisd logger - Migrated to aliquis_hosts (2) - Marching-square: changed default policy from 4-nn to 8-nn in connected components finding - Migrated to aliquis_hosts - Added param ignore_ximage_meta to SOURCE_IMAGE stage - Added custom CRF stage in tools/crf/ - Added bash script for QI IR dataset creation - Modified aliquis_registrtation - Modified aliquis_registration for automatic bitmask production - Refactor of co-registering tool

## 2.0.0 [2018-09-14] - Fix for 16-bit images in SOURCE_IMAGE stage - XImage submodule sync - Script for transform a folder of images with a pre-evaluated transform matrix - Added possibility to give a custom test apl as parameter - Modified tools/model_test/ for SVM model use - Removed black padding in create_dataset for a problem - Modified script for create dataset: black instead of reflect padding - Fix in cmake install script

Aliquis®, Laira®, and Bioretics® are registered trademarks of Bioretics srl.