This is an introduction to「DeepSort」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
DeepSort is a machine learning model for tracking people, assigning IDs to each person.
Traditionally, tracking has used an algorithm called Sort (Simple Online and Realtime Tracking), which uses the Kalman filter. …
This is an introduction to「CenterNet」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
CenterNet is a machine learning model for anchorless object detection published in April 2019.
CenterNet can be used to calculate the bounding boxes for 80 categories of the COCO dataset.
By using heatmaps, as in other systems such as OpenPose, for object detection, CenterNet can perform detection without using anchors used in YOLOv2 and later.
About anchors
An anchor is a bounding box, defined by…
This is an introduction to「FLAVR」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
FLAVR is a machine learning model released in December 2020 that can increase the frame rate of an input video by adding frames.
Architecture
Frame generation relies on video frame interpolation, and is generally used in the following three ways
・Phase based
・Flow based
・Kernel based
In the Phase based method, each frame is viewed as a linear combination of wavelets, and the phase…
This is an introduction to「CrowdCounting」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
CrowdCountCascadedMtl is a machine learning model released in August 2017 that counts the number of people in an input image. It is suitable for counting attendance in large crowds such as concert halls or stadiums for example.
Architecture
CrowdCounting calculates a DensityMap that shows the distribution of the crowd, and predicts the number of people by making estimation based on this Density Map.
This is a tutorial on compressing and obfuscating machine learning models usin the ailia SDK, a cross-platform GPU-enabled fast AI inference framework. More information about ailia SDK can be found here.
Compression of machine learning models
Machine learning models tend to be large in size. For example, ResNet50 is 102.7MB, which puts pressure on communication lines and storage.
The ailia SDK has the ability to compress machine learning models to roughly 1/3 of its original size.
As a parameter for compression, the number of bits can be specified in the range of 16 to 12. …
This is an introduction to「GAST」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
GAST (A Graph Attention Spatio-temporal Convolutional Network for 3D Human Pose Estimation in Video) is a model for predicting 3D skeletons from 2D skeletons that was released in October 2020.
Architecture
GAST takes a time series of 2D skeletons as input and outputs 3D skeletons. YOLOv3 and pose_hrnet_w48_384x288 are used for 2D skeleton detection. …
This is an introduction to「UnetSourceSeparation」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
UnetSourceSeparation is a audio separation model released in March 2019. It can cancel background noise from an input audio file and extract voices.
UnetSourceSeparation demonstration
The official demo of voice separation is shown below. The original voice and the processed voice are played alternately.
Architecture
In speech processing, it is common to perform Short Time Fourier Transform (STFT) on the input speech and apply CNN…
We are pleased to introduce version 1.2.7 of ailia SDK, a cross-platform framework to perform fast AI inference on GPU or CPU. You can find more information about ailia SDK on the official website.
ailia SDK 1.2.7 is a release that focuses on speeding up the inference process.
Performance increase on CPU
ConvTransposeND has benefited a substantial speed up. In particular, the performance of the audio processing system has greatly improved. The CPU performance on Mac M1 has also increased overall.
Performance increase on GPU
In Eltwise, a dedicated process for Tensor Broadcast has been added. The speedup is even more significant with EfficientNet. Also, GPU support for…
This is an introduction to「CodesForLaneDetection」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
CodesForLaneDetection is a machine learning model released in August 2019. This model segments white lines on the road at the pixel level for an input image and can be used for applications such as automated driving.
This is an introduction to「Deep Image Matting」, a machine learning model that can be used with ailia SDK. You can easily use this model to create AI applications using ailia SDK as well as many other ready-to-use ailia MODELS.
Overview
Image matting refers to the problem of extracting interesting targets, usually objects in the foreground, from a static image or a video sequence, which has played an important role in many image and video editing applications.
Deep Image Matting is a machine learning model to perform highly accurate foreground estimation that was announced in April 2017.
Deep Image Matting performs foreground…
