13.2. Release notes version 08.05.00¶
13.2.1. New in this Release¶
Added DL-Inferer GStreamer plugin which can be called from gst-launch command
Added DL-postproc GStreamer plugin which handles text overlay, drawing boxes, lines etc. natively in NV12 format
Added perfOverlay GStreamer plugin which overlays CPU load, HWA load, DDR bandwidth on the post-processed image
Added multi platform support which detects underlying platform at runtime
Added split-output support in edgeai-gst-apps to channel post-processed output to both encode and display
DL-Inferer component in edgeai-gst-apps is migrated to separate repo supporting TFLite RT, ONNX-RT and NeoAI-DLR
13.2.2. Fixed in this Release¶
EDGEAI_ROBOTICS-818 - TVM Models when used with file input does not exit cleanly
EDGEAI_ROBOTICS-814 - addr[0] is INVALID error reported by edgeai-gst-apps infrequently
EDGEAI_ROBOTICS-730 - Performance Visualization Tool command typo/error
EDGEAI_ROBOTICS-666 - No valid frames found before end of stream for image file input and video file output
EDGEAI_ROBOTICS-639 - tiovx mosaic plugin throws error if the pixel ratio of input videos is different
EDGEAI_ROBOTICS-335 - Issue with running setup script on HOST due to ldconfig
13.2.3. Known Issues¶
EDGEAI_ROBOTICS-820 - Video container formats of .mp4 and .mov fails while writing in Docker
EDGEAI_ROBOTICS-819 - tiovxisp element infrequently throws error while exiting edgeai-gst-apps
EDGEAI_ROBOTICS-648 - tiovxdlcolorconvert fails when the image height/2 is not an even number
EDGEAI_ROBOTICS-642 - Mosaic element hangs when both instances of MSC are used in multi-input multi-inference usecase
EDGEAI_ROBOTICS-640 - LDC introduces image format while working with 1936 x 1096 resolution
13.2.4. Model Zoo¶
Models from edgeai-modelzoo are regressed with multiple input sources such as camera and video files. A list of unsupported models are as below,
- Height/2 is odd, tiovxdlpreproc block fetch logic fails
TFL-CL-0100-efficientNet-edgeTPU-m
TFL-CL-0140-efficientNet-lite4
TFL-CL-0170-efficientNet-lite1
TFL-CL-0190-efficientNet-edgeTPU-l
- High latency causes DSS underflow
ONR-SS-8730-deeplabv3-mobv3-lite-large-cocoseg21-512x512
- Odd resolution, tiovxdlcolorconvert fails
TVM-CL-3430-gluoncv-mxnet-xception
TFL-CL-0040-InceptionNetV3
- Human pose model with unsupported post processing
ONR-KD-7000-human-pose-ae-mobv2-fpn-spp-udp-512x512
ONR-KD-7010-human-pose-ae-res50v2-fpn-spp-udp-512x512
ONR-KD-7020-human-pose-ae-mobv2-pan-spp-udp-512x512
ONR-KD-7030-human-pose-ae-res50v2-pan-spp-udp-512x512
Note
Please also refer to edgeai-gst-plugins issues for more details on bugs filed on custom GStreamer plugins
13.2.5. Software components¶
List of software components used in this version
Componenet |
Version |
|---|---|
PSDKLA (filesystem, uboot) |
08.05.00.08 |
PSDKRA (firmware) |
08.05.00.11 |
Docker |
19.03.8-ce |
Ubuntu (Docker) |
20.04.LTS |
Python |
3.8 |
OpenCV |
4.1.0 |
GStreamer |
1.16.3 |
Cmake |
3.16.5 |
Ninja |
1.10.0 |
Meson |
0.53.2 |
Jupyterlab |
3.0.14 |
NeoAI - DLR |
1.10.0 |
Tensorflow |
TIDL_PSDK_8.5 |
TFLite-runtime |
TIDL_PSDK_8.5 |
ONNX-runtime |
TIDL_PSDK_8.5 |
PyYAML |
5.3.1 |
TI Model Zoo |
8.5.0 |
edgeai-gst-plugins |
0.8.0 |
edgeai-tiovx-modules |
8.5 |
edgeai-dl-inferer |
8.5 |
edgeai-tidl-tools |
08_05_00_11 |