動作確認済みモデル

• 下記のモデルはiinferで推論実行できることを確認しています。

• なお、以下に示すサイトの事前学習モデルは、配布元のライセンスに従って使用してください。

Segmentation

header: "base"	frameWork	input	config	model	predict_type
PSPNet_r18(mmseg)	mmsegmentation	512x1024	pspnet_r18-d8_4xb2-80k_cityscapes-512x1024.py	pspnet_r18-d8	mmseg_seg_PSPNet
PSPNet_r50(mmseg)	mmsegmentation	512x1024	pspnet_r50-d8_4xb2-80k_cityscapes-512x1024.py	pspnet_r50-d8	mmseg_seg_PSPNet
PSPNet_r101(mmseg)	mmsegmentation	512x1024	pspnet_r101-d8_4xb2-80k_cityscapes-512x1024.py	pspnet_r101-d8	mmseg_seg_PSPNet
Swin-T(mmseg)	mmsegmentation	512x512	swin-tiny-patch4-window7-in1k-pre_upernet_8xb2-160k_ade20k-512x512.py	upernet_swin_tiny	mmseg_seg_SwinUpernet
Swin-S(mmseg)	mmsegmentation	512x512	swin-small-patch4-window7-in1k-pre_upernet_8xb2-160k_ade20k-512x512.py	upernet_swin_small	mmseg_seg_SwinUpernet
Swin-B(mmseg)	mmsegmentation	512x512	swin-base-patch4-window12-in22k-384x384-pre_upernet_8xb2-160k_ade20k-512x512.py	upernet_swin_base	mmseg_seg_SwinUpernet
SAN(mmseg)	mmsegmentation	640x640	san-vit-b16_coco-stuff164k-640x640.py	san-vit-b16	mmseg_seg_San
SAN(mmseg)	mmsegmentation	640x640	san-vit-l14_coco-stuff164k-640x640.py	san-vit-l14	mmseg_seg_San

Object Detection

header: "base"	frameWork	input	config	model	predict_type
YOLOX(mmdet)	mmdetection	416x416	yolox_tiny_8xb8-300e_coco.py	YOLOX-tiny	mmdet_det_YoloX_Lite
YOLOX(mmdet)	mmdetection	640x640	yolox_s_8xb8-300e_coco.py	YOLOX-s	mmdet_det_YoloX
YOLOX(mmdet)	mmdetection	640x640	yolox_l_8xb8-300e_coco.py	YOLOX-l	mmdet_det_YoloX
YOLOX(mmdet)	mmdetection	640x640	yolox_x_8xb8-300e_coco.py	YOLOX-x	mmdet_det_YoloX
YOLOX	onnx	416x416	yolox_nano.py	ONNX-YOLOX-Nano※1	onnx_det_YoloX_Lite
YOLOX	onnx	416x416	yolox_tiny.py	ONNX-YOLOX-Tiny※1	onnx_det_YoloX_Lite
YOLOX	onnx	640x640	yolox_s.py	ONNX-YOLOX-s※1	onnx_det_YoloX
YOLOX	onnx	640x640	yolox_m.py	ONNX-YOLOX-m※1	onnx_det_YoloX
YOLOX	onnx	640x640	yolox_l.p	ONNX-YOLOX-l※1	onnx_det_YoloX
YOLOX	onnx	640x640	yolox_x.py	ONNX-YOLOX-x※1	onnx_det_YoloX
YOLOv3	onnx	416x416		ONNX-YOLOv3-10	onnx_det_YoloV3
YOLOv3	onnx	416x416		ONNX-YOLOv3-12	onnx_det_YoloV3
YOLOv3	onnx	416x416		ONNX-YOLOv3-12-int8	onnx_det_YoloV3
TinyYOLOv3	onnx	416x416		ONNX-TinyYOLOv3	onnx_det_TinyYoloV3

• ※1 : pth2onnx を使用してONNX形式に変換して使用*

Image Classification

header: "base"	frameWork	input	config	model	predict_type
Swin Transformer	mmpretrain	224x224	swin-tiny_16xb64_in1k.py	swin-tiny_16xb64_in1k	mmpretrain_cls_swin_Lite
Swin Transformer	mmpretrain	224x224	swin-small_16xb64_in1k.py	swin-small_16xb64_in1k	mmpretrain_cls_swin_Lite
Swin Transformer	mmpretrain	384x384	swin-base_16xb64_in1k-384px.py	swin-base_16xb64_in1k-384px	mmpretrain_cls_swin
Swin Transformer	mmpretrain	384x384	swin-large_16xb64_in1k-384px.py	swin-large_16xb64_in1k-384px	mmpretrain_cls_swin
EfficientNet-Lite4	onnx	224x224		EfficientNet-Lite4-11	onnx_cls_EfficientNet_Lite4
EfficientNet-Lite4	onnx	224x224		EfficientNet-Lite4-11-int8	onnx_cls_EfficientNet_Lite4
EfficientNet-Lite4	onnx	224x224		EfficientNet-Lite4-11-qdq	onnx_cls_EfficientNet_Lite4

Face Detection and Recognition

header: "base"	frameWork	input	config	model	predict_type
insightface	insightface	640x640		antelopev2	insightface_det
insightface	insightface	640x640		buffalo_l	insightface_det
insightface	insightface	640x640		buffalo_m	insightface_det
insightface	insightface	640x640		buffalo_s	insightface_det
insightface	insightface	640x640		buffalo_sc	insightface_det