updated to v3.0

Added notice on tracking disabling and upgraded instructions to support v3.0 through python3.9

README.md

@@ -15,6 +15,8 @@ PyAFAR GPU capabilities work only on Linux or WSL2.
 - `AU predictions`: Normalized faces are used for AU predictions (occurrence and intensity). Separate detection modules for occurrence are available for adults and infants. Intensity predictions are available for adults only.
 - `Output`: PyAFAR can output frame-level predictions in CSV and JSON formats to enable easy reading with most platforms used by both computational as well as domain experts.
 
+As MediaPipe retired its support for python 3.7, PyAFAR has disabled person tracking. We are working a fix.
+
 ### For installation instructions click [here](https://github.com/AffectAnalysisGroup/PyAFAR/wiki/3.-Installation) and how to use click [here](https://github.com/AffectAnalysisGroup/PyAFAR/wiki/4.-How-to-use).
 
 ## Citations

index.html

@@ -63,6 +63,12 @@ <h6>
         </div>
     </div>
   </section>
+
+    <section>
+    <div class="col-12 text-center">
+    <h5><font color="red">PyAFAR has disabled person tracking as MediaPipe dependency no longer supports python 3.7. A fix will be issued soon.</font></h5>
+  </div>
+  </section>
 
   <section>
     <div class="jumbotron-grey">

resources/readme/instructions.md

@@ -21,6 +21,7 @@ These are necessary settings that include path to videos to process, save and ot
 - `Use GPU`: GPUs enable faster AU predictions. If your machine has a GPU that you wish to use, enable this.
 - `Batch size`: The batch size for AU predictions. Larger GPUs memory can accommodate higher batch size.
 - `Preprocessing batch size`: Batch size parameter for all pre-AU prediction operations. A higher batch size would require more RAM but speeds up the pipeline.
-- `Person Tracking`: Enables individual tracking in case of a multiperson video. Recommended to disable it if the video contains one individual to speed up the pipeline.
 - `AU occurrence`: Choice of occurrence predictions required from the Input Videos.
 - `AU intensity`: Choice of intensity predictions required from the Input Videos.
+
+The `Person Tracking` feature that tracks individuals in case of a multiperson video, is now disabled as MediaPipe deprecated python 3.7 support. We are working on the fix.