fix: classification profile to use efficientnetb0_FP32INT8 config mod…

…el (intel-retail#433) * fix: classification profile to use efficientnetb0_FP32INT8 config model closes: intel-retail#420 Signed-off-by: Valina Li <[email protected]>
jim-wang-intel · Jan 3, 2024 · 6e0e736 · 6e0e736
1 parent f126b81
commit 6e0e736
Show file tree

Hide file tree

Showing 4 changed files with 2 additions and 16 deletions.
diff --git a/configs/opencv-ovms/envs/classification.env b/configs/opencv-ovms/envs/classification.env
@@ -1,4 +1,4 @@
-CLASSIFICATION_MODEL_NAME=efficientnet-b0
+CLASSIFICATION_MODEL_NAME=efficientnetb0_FP32INT8
 CLASSIFICATION_LABEL_FILE=imagenet_2012.txt
 CLASSIFICATION_OUTPUT_RESOLUTION=1920x1080
 MQTT=

diff --git a/configs/opencv-ovms/models/2022/config_template.json b/configs/opencv-ovms/models/2022/config_template.json
@@ -48,16 +48,6 @@
         "target_device": "{target_device}"
       }
     },
-    {
-      "config": {
-        "name": "efficientnet-b0",
-        "base_path": "/models/efficientnet-b0/FP32-INT8",
-        "target_device": "{target_device}"
-      },
-      "latest": {
-        "num_versions": 1
-      }
-    },
     {
       "config": {
         "name": "ssd_mobilenet_v1_coco",

diff --git a/docker-compose.yml b/docker-compose.yml
@@ -158,7 +158,7 @@ services:
       - GRPC_PORT=9001
       - INPUTSRC=rtsp://localhost:8554/camera_0
       - cid_count=2
-      - CLASSIFICATION_MODEL_NAME=efficientnet-b0
+      - CLASSIFICATION_MODEL_NAME=efficientnetb0_FP32INT8
       - CLASSIFICATION_LABEL_FILE=imagenet_2012.txt
       - CLASSIFICATION_OUTPUT_RESOLUTION=1280x720
       - RENDER_MODE=1 #RENDER_MODE=1 will work only after running xhost +local:docker

diff --git a/docs_src/OVMS/capiYolov5EnsemblePipelineRun.md b/docs_src/OVMS/capiYolov5EnsemblePipelineRun.md
@@ -17,10 +17,6 @@ The model template configuration file has been updated with model configs of yol
 !!! Note
     New model yolov5 is similar to yolov5s configuration except the layout difference.
 
-!!! Note
-    The model efficientnetb0_FP32INT8 is different model from efficientnet-b0.
-
-
 ## Add pipeline specific files
 
 The pre-processing and post-processing work files are added in directory of [`configs/opencv-ovms/gst_capi/pipelines/capi_yolov5_ensemble/`](https://github.com/intel-retail/automated-self-checkout/blob/main/configs/opencv-ovms/gst_capi/pipelines/capi_yolov5_ensemble/), please view directory for details.