zsx

README.md

@@ -1,4 +1,35 @@
 # Hierarchical Object-to-Zone Graph for Object Navigation
-## Setup
-----
+Sixian Zhang, Xinhang Song, Yubing Bai, Weijie Li, Yakui Chu, Shuqiang Jiang (Accepted by ICCV 2021)
 
+[ICCV 2021 Paper]() | 
+[Arxiv Paper](https://arxiv.org/abs/2109.02066) | 
+[Video demo](https://drive.google.com/file/d/1UtTcFRhFZLkqgalKom6_9GpQmsJfXAZC/view)
+## Setup
+- Clone the repository `git clone https://github.com/sx-zhang/HOZ.git` and move into the top-level directory `cd HOZ`
+- Install the dependencies. `pip install -r requirements.txt`
+- We provide pre-trained model of [hoz](https://drive.google.com/file/d/11L-ejoWgLHPBe_F-gQ7dJ5gQZB0dzNjr/view?usp=sharing) and [hoztpn](). For evaluation and fine-tuning training, you can download them to the `trained_models` directory.
+- Download the [dataset](https://drive.google.com/file/d/1kvYvutjqc6SLEO65yQjo8AuU85voT5sC/view), which refers to [ECCV-VN](https://github.com/xiaobaishu0097/ECCV-VN). The offline data is discretized from [AI2THOR](https://ai2thor.allenai.org/) simulator.  
+  Your `data` folder should look like this
+
+  `Data 
+    └── AI2thor_offline_data/
+    ├── FloorPlan1
+    │   ├── resnet18_featuremap.hdf5
+    │   ├── graph.json
+    │   ├── visible_object_map_1.5.json
+    │   ├── det_feature_categories.hdf5
+    │   ├── grid.json
+    │   └── optimal_action.json
+    ├── FloorPlan2
+    └── ...`
+## HOZ graph Construction
+## Training and Evaluation
+### Train the baseline model 
+`python main.py --title Basemodel --model BaseModel --workers 12 –gpu-ids 0`
+### Train our HOZ model 
+`python main.py --title HOZ --model HOZ --workers 12 –gpu-ids 0`
+### Evaluate our HOZ model 
+`python python full_eval.py --title HOZ --model HOZ --results_json HOZ.json --gpu-ids 0`
+### Evaluate our HOZ-TPN model 
+`python python full_eval.py --title HOZTPN --model MetaMemoryHOZ --results_json HOZTPN.json --gpu-ids 0`
+## Citing

utils/command_parser.py

@@ -350,7 +350,7 @@ def parse_arguments():
     parser.add_argument(
         '--data-dir',
         type=str,
-        default='/tmp_data/Scene_Data',
+        default='./datasets/Scene_Data',
         help='where dataset is stored.',
     )