Update README.md

README.md

@@ -32,16 +32,20 @@ download_huggingface_model("superanimal_quadruped", model_dir)
 ## Intended Use
 • Intended to be used for pose estimation of quadruped images taken from side-view. The model serves a better starting
 point than ImageNet weights in downstream datasets such as AP-10K.
+
 • Intended for academic and research professionals working in fields related to animal behavior, such as neuroscience
 and ecology.
+
 • Not suitable as a zeros-shot model for applications that require high keypiont precision, but can be fine-tuned with
 minimal data to reach human-level accuracy. Also not suitable for videos that look dramatically different from those
 we show in the paper.
-Factors
+
+## Factors
+
 • Based on the known robustness issues of neural networks, the relevant factors include the lighting, contrast and
 resolution of the video frames. The present of objects might also cause false detections and erroneous keypoints.
 When two or more animals are extremely close, it could cause the top-down detectors to only detect only one animal,
-if used without further fine-tuning or with a method such as BUCTD (36).
+if used without further fine-tuning or with a method such as BUCTD (Zhou et al. 2023 ICCV).
 
 ## Metrics
 • Mean Average Precision (mAP)
@@ -78,12 +82,6 @@ Here is an image with the keypoint guide:
 <img src="https://images.squarespace-cdn.com/content/v1/57f6d51c9f74566f55ecf271/1690988780004-AG00N6OU1R21MZ0AU9RE/modelcard-SAQ.png?format=1500w" width="95%">
 </p>
 
-Please note that each dataset was labeled by separate labs \& separate individuals, therefore while we map names
-to a unified pose vocabulary (found here: https://github.com/AdaptiveMotorControlLab/modelzoo-figures), there will be annotator bias in keypoint placement (See the Supplementary Note on annotator bias). 
-You will also note the dataset is highly diverse across species, but collectively has more representation of domesticated animals like dogs, cats, horses, and cattle. 
-We recommend if performance is not as good as you need it to be, first try video adaptation (see Ye et al. 2023), 
-or fine-tune these weights with your own labeling.
-
 
 ## Ethical Considerations
 
@@ -96,8 +94,12 @@ characteristics not well-represented in the training data.
 
 • Please note that each dataest was labeled by separate labs & separate individuals, therefore while we map names to a
 unified pose vocabulary, there will be annotator bias in keypoint placement (See Ye et al. 2023 for our Supplementary
-Note on annotator bias). You will also note the dataset is highly diverse across species, but collectively has more
-representation of domesticated animals like dogs, cats, horses, and cattle. We recommend if performance is not as
+Note on annotator bias). 
+
+• Note the dataset is highly diverse across species, but collectively has more
+representation of domesticated animals like dogs, cats, horses, and cattle. 
+
+• We recommend if performance is not as
 good as you need it to be, first try video adaptation (see Ye et al. 2023), or fine-tune these weights with your own
 labeling.