Merge pull request #165 from LokeZhou/deploy

Deploy

Author: LokeZhou

deploy/groundingdino/README.md

@@ -2,26 +2,26 @@
 
 ## 1. 模型简介
 
-Paddle implementation of [Grounding DINO](https://arxiv.org/abs/2303.05499), a stronger open-set object detector.
+该模型是 [Grounding DINO](https://arxiv.org/abs/2303.05499) 的 paddle 实现。
 
 
-## 2. Demo
+## 2. 示例
 
-## 2.1 prepare
+## 2.1 依赖安装 （可选）
 ```bash
 #Multi-scale deformable attention custom OP compilation
 cd /paddlemix/models/groundingdino/csrc/
 python setup_ms_deformable_attn_op.py install
 
 ```
-## 2.2 Export model for static inference
+## 2.2 静态图导出与预测
 ```bash
-#export grounding dino model
+#静态图模型导出
 python export.py \
 --dino_type GroundingDino/groundingdino-swint-ogc
 
 
-#inference
+#静态图预测
  python predict.py  \
  --text_encoder_type GroundingDino/groundingdino-swint-ogc \
  --model_path output_groundingdino/GroundingDino/groundingdino-swint-ogc \

deploy/sam/README.md

@@ -1,33 +1,35 @@
-#
+# Segment Anything
 
 ## 1. 模型简介
 
-Paddle implementation of [Segment Anything](https://ai.facebook.com/research/publications/segment-anything/), produces high quality object masks from input prompts such as points or boxes.
+该模型是 [Segment Anything](https://ai.facebook.com/research/publications/segment-anything/) 的 paddle 实现, 可输入点或框进行分割。
 
 
-## 2. Demo
+## 2. 示例
 
-## 2.2 Export model for static inference
+## 2.1 静态图导出与预测
 ```bash
-#export sam model input_type box
+#导出输入类型是 bbox 的静态图
 python export.py --model_type Sam/SamVitH-1024 --input_type boxs  --save_dir sam_export
 
-#export sam model input_type points
+#导出输入类型是 points 的静态图
 python export.py --model_type Sam/SamVitH-1024 --input_type points  --save_dir sam_export
 
 
 
-#boxs prompt
+#bbox 提示词推理
 python predict.py
 --input_image image_you_want_to_detect.jpg \
 --box_prompt 548 372 593 429 443 374 482 418 \
 --input_type boxs \
+--model_name_or_path Sam/SamVitH-1024 \
 --cfg sam_export_SamVitH_boxs/deploy.yaml
 
-#points prompt
+#points 提示词推理
 python predict.py \
 --input_image mage_you_want_to_detect.jpg \
 --points_prompt 548 372 \
 --input_type points \
+--model_name_or_path Sam/SamVitH-1024 \
 --cfg sam_export_SamVitH_points/deploy.yaml
 ```

deploy/sam/export.py

@@ -26,7 +26,7 @@ def parse_args():
     parser = argparse.ArgumentParser(description="Export Inference Model.")
     parser.add_argument(
         "--model_type",
-        choices=["SamVitL", "SamVitB", "SamVitH"],
+        choices=["Sam/SamVitH-1024", "Sam/SamVitB", "Sam/SamVitL"],
         required=True,
         help="The model type.",
         type=str,

deploy/sam/predict.py

@@ -244,7 +244,7 @@ class ModelArguments:
     """
 
     model_name_or_path: str = field(
-        default="Sam/SamVitH",
+        default="Sam/SamVitH-1024",
         metadata={"help": "Path to pretrained model or model identifier"},
     )
     input_type: str = field(

paddlemix/examples/Sam/README.md

@@ -1,13 +1,14 @@
-#
+# Segment Anything
 
 ## 1. 模型简介
 
-Paddle implementation of [Segment Anything](https://ai.facebook.com/research/publications/segment-anything/), produces high quality object masks from input prompts such as points or boxes.
+该模型是 [Segment Anything](https://ai.facebook.com/research/publications/segment-anything/) 的 paddle 实现, 可输入点或框进行分割。
 
 
-## 2. Demo
 
-## 2.1 dynamic inference
+## 2. 示例
+
+## 2.1 动态图推理
 ```bash
 #box
 python run_predict.py \

paddlemix/examples/groundingdino/README.md

@@ -2,19 +2,19 @@
 
 ## 1. 模型简介
 
-Paddle implementation of [Grounding DINO](https://arxiv.org/abs/2303.05499), a stronger open-set object detector.
+该模型是 [Grounding DINO](https://arxiv.org/abs/2303.05499) 的 paddle 实现。
 
 
 ## 2. Demo
 
-## 2.1 prepare
+## 2.1 依赖安装（可选）
 ```bash
 #Multi-scale deformable attention custom OP compilation
 cd /paddlemix/models/groundingdino/csrc/
 python setup_ms_deformable_attn_op.py install
 
 ```
-## 2.2 dynamic inference
+## 2.2 动态图推理
 ```bash
 python run_predict.py \
 --input_image image_you_want_to_detect.jpg \