| Application | Network | TensorRT | Open
VINO |
Onnx
Runtime |
Translator
Plugin |
Minimun required EVA |
Reference
link |
Note |
| Classification | GoogLeNet | v | v | x | adtrans_classifier | 1.0+ | bvlc_googlenet |
|
| MobilenetV2 | v | v | v | adtrans_classifier | 2.0+ | mobilenetv2-7.onnx |
|
|
| ResNet18 | v | x | x | adtrans_classifier | 3.5.3+ | tlt_resnet18_ssd |
|
|
| ResNet | - | - | v | adtrans_classifier | 3.5.3+ | resnet50-v1-7.onnx |
|
|
| SqueezeNet | - | - | v | adtrans_classifier | 3.5.3+ | squeezenet1.0-7.onnx |
|
|
| VGG | - | - | v | adtrans_classifier | 3.5.3+ | vgg16-7.onnx |
|
|
| AlexNet | - | - | v | adtrans_classifier | 3.5.3+ | bvlcalexnet-7.onnx |
|
|
| GoogleNet | - | - | v | adtrans_classifier | 3.5.3+ | googlenet-7.onnx |
|
|
| CaffeNet | - | - | v | adtrans_classifier | 3.5.3+ | caffenet-7.onnx |
|
|
| RCNN_ILSVRC13 | - | - | v | adtrans_classifier | 3.5.3+ | rcnn-ilsvrc13-7.onnx |
|
|
| DenseNet-121 | - | - | v | adtrans_classifier | 3.5.3+ | densenet-7.onnx |
|
|
| Inception_V1/V2 | - | - | v | adtrans_classifier |
3.5.3+ |
inception-v1-7.onnx /
inception-v2-7.onnx |
|
|
| ZFNet-512 | - | - | v | adtrans_classifier | 3.5.3+ | zfnet512-7.onnx |
|
|
| ShuffleNet_V1/V2 | - | - | v | adtrans_classifier | 3.5.3+ | shufflenet-7.onnx /
shufflenet-v2-10.onnx |
|
|
| EfficientNet-Lite4 | - | - | v | adtrans_classifier | 3.5.3+ | efficientnet-lite4-11.onnx |
|
|
| Object Detection | SSD Inception-V2 | v | - | x | adtrans_ssd | 2.0+ | sampleUffSSD |
|
| Face detection
(SSD based) |
x | v | x | adtrans_ssd | 1.0+ | face_detection |
|
|
| Yolov3 | v | v | v | adtrans_yolo | 2.0+ | AlexeyAB/darknet
yolov3_onnx |
* | |
| Yolov4 | v | v | v | adtrans_yolo | 3.5.3+ | AlexeyAB/darknet
yolov3_onnx |
* | |
| Yolov3-tiny | v | v | v | adtrans_yolo | 3.5.3+ | AlexeyAB/darknet
yolov3_onnx |
* | |
| Yolov4-tiny | v | v | v | adtrans_yolo | 3.5.3+ | AlexeyAB/darknet
yolov3_onnx |
(1) | |
| Yolov2-tiny | - | - | v | adtrans_tinyyolov2_py | 3.5.3+ | CustomVision | (2) | |
| Pose Estimation | OpenPose | v | v | v | adtrans_openpose | 3.0.0+ | lightweight-human-pose-estimation | (3) |
| PoseNet | v | v | v | adtrans_posenet | 3.8.1+ | deepstream_pose_estimation | (4) | |
| Segmentation | FCN-Alexnet | v | x | - | adtrans_fcn | 2.0+ | L4T-R32.2#pre-trained-models |
|
| Intel Segmentation | x | v | - | adtrans_segment | 2.0+ | semantic_segmentation |
|
(1) : May need to modify conversion script
(2) : Only for CustomVision object detection General Compact Domain
(3) : Better in crowed and small size people
(4) : Better for single people
V : Tested , convert OK
- : Not test, possible to optimize
X : Tested , cannot optimize