final renaming changes

README.md

@@ -18,10 +18,12 @@ pip install -e .[reality]
 ```
 Install all the dependencies:
 ```bash
-git clone git@github.com:WongKinYiu/yolov7.git  # if using YOLOv7
 git clone git@github.com:IDEA-Research/GroundingDINO.git
+git clone git@github.com:WongKinYiu/yolov7.git  # if using YOLOv7
 ```
-Follow the install directions for GroundingDINO. Nothing needs to be done for YOLOv7, but it needs to be cloned into the repo.
+Follow the original install directions for GroundingDINO, which can be found here: https://github.com/IDEA-Research/GroundingDINO.
+
+Nothing needs to be done for YOLOv7, but it needs to be cloned into the repo.
 
 ### Installing GroundingDINO (Only if using conda-installed CUDA)
 Only attempt if the installation instructions in the GroundingDINO repo do not work.

config/experiments/reality.yaml

@@ -3,7 +3,7 @@
 # Copyright (c) 2023 Boston Dynamics AI Institute LLC. All rights reserved.
 
 defaults:
-  - /policy: zsos_config_base
+  - /policy: vlfm_config_base
   - _self_
 
 policy:

config/experiments/vlfm_objectnav_hm3d.yaml

@@ -14,7 +14,7 @@ defaults:
   - /habitat/task/measurements:
     - frontier_exploration_map
     - traveled_stairs
-  - /habitat_baselines/rl/policy: zsos_policy
+  - /habitat_baselines/rl/policy: vlfm_policy
   - _self_
 
 habitat:

vlfm/policy/base_objectnav_policy.py

@@ -406,7 +406,7 @@ def _infer_depth(
 
 
 @dataclass
-class ZSOSConfig:
+class VLFMConfig:
     name: str = "HabitatITMPolicy"
     text_prompt: str = "Seems like there is a target_object ahead."
     pointnav_policy_path: str = "data/pointnav_weights.pth"
@@ -421,16 +421,16 @@ class ZSOSConfig:
     hole_area_thresh: int = 100000
     use_vqa: bool = False
     vqa_prompt: str = "Is this "
-    coco_threshold: float = 0.6
+    coco_threshold: float = 0.8
     non_coco_threshold: float = 0.4
     agent_radius: float = 0.18
 
     @classmethod  # type: ignore
     @property
     def kwaarg_names(cls) -> List[str]:
         # This returns all the fields listed above, except the name field
-        return [f.name for f in fields(ZSOSConfig) if f.name != "name"]
+        return [f.name for f in fields(VLFMConfig) if f.name != "name"]
 
 
 cs = ConfigStore.instance()
-cs.store(group="policy", name="zsos_config_base", node=ZSOSConfig())
+cs.store(group="policy", name="vlfm_config_base", node=VLFMConfig())

vlfm/policy/habitat_policies.py

@@ -22,7 +22,7 @@
 from vlfm.vlm.grounding_dino import ObjectDetections
 
 from ..mapping.obstacle_map import ObstacleMap
-from .base_objectnav_policy import BaseObjectNavPolicy, ZSOSConfig
+from .base_objectnav_policy import BaseObjectNavPolicy, VLFMConfig
 from .itm_policy import ITMPolicy, ITMPolicyV2, ITMPolicyV3
 
 HM3D_ID_TO_NAME = ["chair", "bed", "potted plant", "toilet", "tv", "couch"]
@@ -95,9 +95,9 @@ def __init__(
     def from_config(
         cls, config: DictConfig, *args_unused: Any, **kwargs_unused: Any
     ) -> "HabitatMixin":
-        policy_config: ZSOSPolicyConfig = config.habitat_baselines.rl.policy
+        policy_config: VLFMPolicyConfig = config.habitat_baselines.rl.policy
         kwargs = {
-            k: policy_config[k] for k in ZSOSPolicyConfig.kwaarg_names  # type: ignore
+            k: policy_config[k] for k in VLFMPolicyConfig.kwaarg_names  # type: ignore
         }
 
         # In habitat, we need the height of the camera to generate the camera transform
@@ -287,9 +287,9 @@ class HabitatITMPolicyV3(HabitatMixin, ITMPolicyV3):
 
 
 @dataclass
-class ZSOSPolicyConfig(ZSOSConfig, PolicyConfig):
+class VLFMPolicyConfig(VLFMConfig, PolicyConfig):
     pass
 
 
 cs = ConfigStore.instance()
-cs.store(group="habitat_baselines/rl/policy", name="zsos_policy", node=ZSOSPolicyConfig)
+cs.store(group="habitat_baselines/rl/policy", name="vlfm_policy", node=VLFMPolicyConfig)

vlfm/policy/reality_policies.py

@@ -10,7 +10,7 @@
 from torch import Tensor
 
 from vlfm.mapping.obstacle_map import ObstacleMap
-from vlfm.policy.base_objectnav_policy import ZSOSConfig
+from vlfm.policy.base_objectnav_policy import VLFMConfig
 from vlfm.policy.itm_policy import ITMPolicyV2
 
 INITIAL_ARM_YAWS = np.deg2rad([-90, -60, -30, 0, 30, 60, 90, 0]).tolist()
@@ -48,8 +48,8 @@ def __init__(
     def from_config(
         cls, config: DictConfig, *args_unused: Any, **kwargs_unused: Any
     ) -> Any:
-        policy_config: ZSOSConfig = config.policy
-        kwargs = {k: policy_config[k] for k in ZSOSConfig.kwaarg_names}  # type: ignore
+        policy_config: VLFMConfig = config.policy
+        kwargs = {k: policy_config[k] for k in VLFMConfig.kwaarg_names}  # type: ignore
 
         return cls(**kwargs)
 
@@ -194,7 +194,7 @@ def _infer_depth(
 
 @dataclass
 class RealityConfig(DictConfig):
-    policy: ZSOSConfig = ZSOSConfig()
+    policy: VLFMConfig = VLFMConfig()
 
 
 class RealityITMPolicyV2(RealityMixin, ITMPolicyV2):

vlfm/semexp_env/eval.py

@@ -21,7 +21,7 @@
 args = get_args()
 args.agent = "vlfm"  # Doesn't really matter as long as it's not "sem_exp"
 args.split = "val"
-args.task_config = "objnav_gibson_zsos.yaml"
+args.task_config = "objnav_gibson_vlfm.yaml"
 # Ensure a random seed
 args.seed = int(time.time() * 1000) % 2**32