From f3b71d69cf7910096ffadfadcda7170a9847cc61 Mon Sep 17 00:00:00 2001 From: Evgeny Izutov Date: Tue, 15 Jan 2019 11:12:29 +0300 Subject: [PATCH] Fixed readme --- README.md | 2 ++ 1 file changed, 2 insertions(+) diff --git a/README.md b/README.md index 2e17a8cf..f939832c 100644 --- a/README.md +++ b/README.md @@ -93,7 +93,9 @@ The train procedure for action detection&recognition model consists of two consi 4. [Conversation to MO-compatible format](#conversation-to-mo-compatible-format) **Note 1**: you can skip the first stage (training PD) and significantly speedup the second stage (training AR) by initializing your model from our distributed `.caffemodel` snapshot: `$CAFFE_ROOT/models/person_detection_action_recognition/stage2__action_regognition/init_weights.caffemodel` + **Note 2**: if you want to change the list of supported actions follow next steps: + 1. Change `ACTION_NAMES_MAP` dictionary in data layer `$CAFFE_ROOT/python/custom_layers/actions_data_layer.py` according your list of actions. 2. Replace fields `valid_action_ids` and `valid_class_ids` to list of you valid action IDs for next layers in `$CAFFE_ROOT/models/person_detection_action_recognition/stage1__person_detector/train.prototxt` and `$CAFFE_ROOT/models/person_detection_action_recognition/stage2__action_regognition/train.prototxt` files: - `data` -- GitLab