update demos (open-mmlab#27)

wjkim81 · Jul 23, 2020 · caa5253 · caa5253
1 parent b72d52f
commit caa5253
Show file tree

Hide file tree

Showing 4 changed files with 19 additions and 17 deletions.
diff --git a/demo/img_demo.py b/demo/img_demo.py
@@ -74,6 +74,7 @@ def main():
         if args.out_img_root == '':
             out_file = None
         else:
+            os.makedirs(args.out_img_root, exist_ok=True)
             out_file = os.path.join(args.out_img_root, f'vis_{i}.jpg')
 
         # show the results

diff --git a/demo/img_demo_with_mmdet.py b/demo/img_demo_with_mmdet.py
@@ -76,6 +76,7 @@ def main():
     if args.out_img_root == '':
         out_file = None
     else:
+        os.makedirs(args.out_img_root, exist_ok=True)
         out_file = os.path.join(args.out_img_root, f'vis_{args.img}')
 
     # show the results

diff --git a/demo/video_demo_with_mmdet.py b/demo/video_demo_with_mmdet.py
@@ -57,10 +57,11 @@ def main():
 
     cap = cv2.VideoCapture(args.video_path)
 
-    if args.out_video_root != '':
-        save_out_video = True
-    else:
+    if args.out_video_root == '':
         save_out_video = False
+    else:
+        os.makedirs(args.out_video_root, exist_ok=True)
+        save_out_video = True
 
     if save_out_video:
         fps = cap.get(cv2.CAP_PROP_FPS)

diff --git a/docs/getting_started.md b/docs/getting_started.md
@@ -132,8 +132,8 @@ We provide a demo script to test a single image, given gt json file.
 python demo/image_demo.py \
     ${MMPOSE_CONFIG_FILE} ${MMPOSE_CHECKPOINT_FILE} \
     --img-root ${IMG_ROOT} --json-file ${JSON_FILE} \
-    --show \
-    [--out-img-root ${OUTPUT_DIR} --device ${GPU_ID}] \
+    --out-img-root ${OUTPUT_DIR} \
+    [--show --device ${GPU_ID}] \
     [--kpt-thr ${KPT_SCORE_THR}]
 ```
 
@@ -144,8 +144,7 @@ python demo/image_demo.py \
     configs/top_down/hrnet/coco/hrnet_w48_coco_256x192.py \
     hrnet_w48_coco_256x192/epoch_210.pth \
     --img-root tests/data/ --json-file tests/data/test_coco.json \
-    --show \
-    --out-img-root ./
+    --out-img-root vis_results
 ```
 
 #### Using mmdet for human bounding box detection
@@ -159,8 +158,8 @@ python demo/image_demo_with_mmdet.py \
     ${MMDET_CONFIG_FILE} ${MMDET_CHECKPOINT_FILE} \
     ${MMPOSE_CONFIG_FILE} ${MMPOSE_CHECKPOINT_FILE} \
     --img-root ${IMG_ROOT} --img ${IMG_FILE} \
-    --show \
-    [--out-img-root ${OUTPUT_DIR} --device ${GPU_ID}] \
+    --out-img-root ${OUTPUT_DIR} \
+    [--show --device ${GPU_ID}] \
     [--bbox-thr ${BBOX_SCORE_THR} --kpt-thr ${KPT_SCORE_THR}]
 ```
 
@@ -173,8 +172,7 @@ python demo/img_demo_with_mmdet.py mmdetection/configs/faster_rcnn/faster_rcnn_r
     hrnet_w48_coco_256x192/epoch_210.pth \
     --img-root tests/data/ \
     --img 000000196141.jpg \
-    --show \
-    --out-img-root ./
+    --out-img-root vis_results
 ```
 
 ### Video demo
@@ -188,8 +186,8 @@ python demo/video_demo_with_mmdet.py \
     ${MMDET_CONFIG_FILE} ${MMDET_CHECKPOINT_FILE} \
     ${MMPOSE_CONFIG_FILE} ${MMPOSE_CHECKPOINT_FILE} \
     --video-path ${VIDEO_FILE} \
-    --show \
-    [--output-video-root ${OUTPUT_VIDEO_ROOT} --device ${GPU_ID}] \
+    --output-video-root ${OUTPUT_VIDEO_ROOT} \
+    [--show --device ${GPU_ID}] \
     [--bbox-thr ${BBOX_SCORE_THR} --kpt-thr ${KPT_SCORE_THR}]
 ```
 
@@ -201,8 +199,7 @@ python demo/video_demo_with_mmdet.py mmdetection/configs/faster_rcnn/faster_rcnn
     configs/top_down/hrnet/coco/hrnet_w48_coco_256x192.py \
     hrnet_w48_coco_256x192/epoch_210.pth \
     --video_path demo/demo_video.mp4 \
-    --show \
-    --output-video-root ./
+    --output-video-root vis_results
 ```
 
 ## Train a model
@@ -264,10 +261,12 @@ If you can run MMPose on a cluster managed with [slurm](https://slurm.schedmd.co
 ./tools/slurm_train.sh ${PARTITION} ${JOB_NAME} ${CONFIG_FILE} ${WORK_DIR} [${GPUS}]
 ```
 
-Here is an example of using 16 GPUs to train ResNet50 on the dev partition in a slurm cluster. (use `GPUS_PER_NODE=8` to specify a single slurm cluster node with 8 GPUs.)
+Here is an example of using 16 GPUs to train ResNet50 on the dev partition in a slurm cluster.
+(Use `GPUS_PER_NODE=8` to specify a single slurm cluster node with 8 GPUs, `CPUS_PER_TASK=2` to use 2 cpus per task.
+Assume that `Test` is a valid ${PARTITION} name.)
 
 ```shell
-GPUS_PER_NODE=8 ./tools/slurm_train.sh configs/top_down/resnet/coco/res50_coco_256x192.py work_dirs/res50_coco_256x192 16
+GPUS_PER_NODE=8 CPUS_PER_TASK=2 ./tools/slurm_train.sh Test res50 configs/top_down/resnet/coco/res50_coco_256x192.py work_dirs/res50_coco_256x192 16
 ```
 
 You can check [slurm_train.sh](../tools/slurm_train.sh) for full arguments and environment variables.