preparing settings for model merge

updated naming in experiments.py
small changes to reporting of classifier.py

Author: JensBlack

DeepLabStream.py

@@ -17,7 +17,7 @@
 import pandas as pd
 import click
 
-from utils.configloader import RESOLUTION, FRAMERATE, OUT_DIR, MODEL,  MULTI_CAM, STACK_FRAMES, \
+from utils.configloader import RESOLUTION, FRAMERATE, OUT_DIR, MODEL_NAME,  MULTI_CAM, STACK_FRAMES, \
     ANIMALS_NUMBER, STREAMS, STREAMING_SOURCE
 from utils.poser import load_deeplabcut, get_pose, find_local_peaks_new, calculate_skeletons
 from utils.plotter import plot_bodyparts, plot_metadata_frame
@@ -722,7 +722,7 @@ def show_benchmark_statistics():
 
     if benchmark_enabled:
         import re
-        short_model = re.split('[-_]', MODEL)
+        short_model = re.split('[-_]', MODEL_NAME)
         short_model = short_model[0] + '_' + short_model[2]
         np.savetxt(f'{OUT_DIR}/{short_model}_framerate_{FRAMERATE}_resolution_{RESOLUTION[0]}_{RESOLUTION[1]}.txt', np.transpose([fps_data, whole_loop_time_data]))
 

experiments/custom/classifier.py

@@ -279,6 +279,8 @@ def bsoid_classifier_pool_run(input_q: mp.Queue,output_q: mp.Queue):
             output_q.put((last_prob,feature_id))
             end_time = time.time()
             print("Classification time: {:.2f} msec".format((end_time-start_time)*1000))
+            print("Feature ID: "+ feature_id)
+
         else:
             pass
 

experiments/custom/experiments.py

@@ -12,17 +12,18 @@
 from collections import Counter
 from experiments.custom.stimulus_process import ClassicProtocolProcess, SimpleProtocolProcess,Timer, ExampleProtocolProcess
 from experiments.custom.triggers import ScreenTrigger, RegionTrigger, OutsideTrigger, DirectionTrigger, SpeedTrigger,\
-    SimbaThresholdBehaviorPoolTrigger, BsoidClassBehaviorTriggerPool
+    SimbaThresholdBehaviorPoolTrigger, BsoidClassBehaviorPoolTrigger
 from utils.plotter import plot_triggers_response
 from utils.analysis import angle_between_vectors
 from experiments.custom.stimulation import show_visual_stim_img,laser_switch
-from experiments.custom.classifier import SimbaClassifier_Process, SimbaProcessPool, BsoidProcessPool
+from experiments.custom.classifier import SimbaProcessPool, BsoidProcessPool
 
 
 from utils.configloader import THRESHOLD, POOL_SIZE
 
+
 """ experimental classification experiment using Simba trained classifiers in a pool"""
-class SimbaBehaviorExperimentPool:
+class SimbaBehaviorPoolExperiment:
     """
     Test experiment for Simba classification
     Simple class to contain all of the experiment properties and includes classification
@@ -130,7 +131,7 @@ def get_info(self):
 
 """ experimental classification experiment using BSOID trained classifiers in a pool"""
 
-class BsoidBehaviorExperimentPool:
+class BsoidBehaviorPoolExperiment:
     """
     Test experiment for Simba classification
     Simple class to contain all of the experiment properties and includes classification
@@ -144,7 +145,7 @@ def __init__(self):
         self._process_pool = BsoidProcessPool(POOL_SIZE)
         #pass classifier to trigger, so that check_skeleton is the only function that passes skeleton
         #initiate in experiment, so that process can be started with start_experiment
-        self._behaviortrigger = BsoidClassBehaviorTriggerPool(target_class= THRESHOLD,
+        self._behaviortrigger = BsoidClassBehaviorPoolTrigger(target_class= THRESHOLD,
                                                               class_process_pool = self._process_pool)
         self._event = None
         #is not fully utilized in this experiment but is usefull to keep for further adaptation

experiments/custom/triggers.py

@@ -636,7 +636,7 @@ def get_time_window_len(self):
         return self._time_window_len
 
 
-class BsoidClassBehaviorTriggerPool:
+class BsoidClassBehaviorPoolTrigger:
     """
     Trigger to check if animal's behavior is classified as specific motif with BSOID trained classifier.
     """

requirements.txt

@@ -1,3 +1,4 @@
+numba
 gpiozero
 pigpio
 pyserial

settings.ini

@@ -8,9 +8,13 @@ CAMERA_SOURCE = 0
 #you can use "camera", "ipwebcam" or "video" to select your input source
 STREAMING_SOURCE = video
 
-[DeepLabCut]
-DLC_PATH = DLC_PATH
-MODEL = MODEL_NAME
+[Pose Estimation]
+#possible origins are: DLC
+MODEL_ORIGIN = ORIGIN
+MODEL_PATH = PATH_TO_MODEL
+MODEL_NAME = MODEL_NAME
+; Currently used to translate pose estimation into postures with bodypart names ... not necessary for classic DLC networks
+ALL_BODYPARTS = bp1, bp2, bp3, bp4
 
 [Experiment]
 #Available parameters are "CUSTOM" and "BASE"
@@ -22,13 +26,14 @@ RECORD_EXP = True
 
 [Classification]
 PATH_TO_CLASSIFIER = CLASSIFIER_PATH
-ALL_BODYPARTS = Ear_left_1, Ear_right_1, Nose_1,Center_1, Lateral_left_1, Lateral_right_1, Tail_base_1, Ear_left_2,Ear_right_2, Nose_2, Center_2, Lateral_left_2, Lateral_right_2, Tail_base_2
 #time window used for feature extraction (currently only works with 15)
 TIME_WINDOW = 15
 #number of parallel classifiers to run, this is dependent on your performance time. You need at least 1 more classifier then your average classification time.
 POOL_SIZE = 4
-#threshold to accept a classification probability as positive detection
+#threshold to accept a classification probability as positive detection (SIMBA + )
 THRESHOLD = 0.5
+# class/category of identified behavior to use as trigger (only used for B-SOID)
+TRIGGER = 5
 #feature extraction currently works with millimeter not px, so be sure to enter the factor (as in simba).
 PIXPERMM = 6.132
 

utils/configloader.py

@@ -27,8 +27,12 @@ def get_script_path():
 adv_cfg_path = os.path.join(os.path.dirname(__file__), 'advanced_settings.ini')
 with open(adv_cfg_path) as adv_cfg_file:
     adv_dsc_config.read_file(adv_cfg_file)
-# DeepLabCut
-deeplabcut_config = dict(dsc_config.items('DeepLabCut'))
+
+#poseestimation
+MODEL_ORIGIN = dsc_config['Pose Estimation'].get('MODEL_ORIGIN')
+MODEL_PATH = dsc_config['Pose Estimation'].get('MODEL_PATH')
+MODEL_NAME = dsc_config['Pose Estimation'].get('MODEL_NAME')
+ALL_BODYPARTS = tuple(part for part in dsc_config['Pose Estimation'].get('ALL_BODYPARTS').split(','))
 
 # Streaming items
 try:
@@ -37,7 +41,7 @@ def get_script_path():
     print('Incorrect resolution in config!\n'
           'Using default value "RESOLUTION = 848, 480"')
     RESOLUTION = (848, 480)
-MODEL = dsc_config['Streaming'].get('MODEL')
+
 FRAMERATE = dsc_config['Streaming'].getint('FRAMERATE')
 OUT_DIR = dsc_config['Streaming'].get('OUTPUT_DIRECTORY')
 CAMERA_SOURCE = dsc_config['Streaming'].get('CAMERA_SOURCE')
@@ -59,7 +63,6 @@ def get_script_path():
 #Classification
 PATH_TO_CLASSIFIER = dsc_config['Classification'].get('PATH_TO_CLASSIFIER')
 PIXPERMM = dsc_config['Classification'].getfloat('PIXPERMM')
-ALL_BODYPARTS = [str(part).strip() for part in dsc_config['Classification'].get('ALL_BODYPARTS').split(',')]
 THRESHOLD = dsc_config['Classification'].getfloat('THRESHOLD')
 POOL_SIZE = dsc_config['Classification'].getint('POOL_SIZE')
 TIME_WINDOW = dsc_config['Classification'].getint('TIME_WINDOW')

utils/poser.py

@@ -16,23 +16,24 @@
 from scipy.ndimage.measurements import label, maximum_position
 from scipy.ndimage.morphology import generate_binary_structure, binary_erosion
 from scipy.ndimage.filters import maximum_filter
-from utils.configloader import deeplabcut_config
+from utils.configloader import  MODEL_ORIGIN, MODEL_NAME, MODEL_PATH
 
-MODEL = deeplabcut_config['model']
-DLC_PATH = deeplabcut_config['dlc_path']
 
 # trying importing functions using deeplabcut module, if DLC 2 is installed correctly
-try:
-    import deeplabcut.pose_estimation_tensorflow.nnet.predict as predict
-    from deeplabcut.pose_estimation_tensorflow.config import load_config
-    models_folder = 'pose_estimation_tensorflow/models/'
-# if not DLC 2 is not installed, try import from DLC 1 the old way
-except ImportError:
-    # adding DLC posing path and loading modules from it
-    sys.path.insert(0, DLC_PATH + "/pose-tensorflow")
-    from config import load_config
-    from nnet import predict
-    models_folder = 'pose-tensorflow/models/'
+if MODEL_ORIGIN == 'DLC':
+    try:
+        import deeplabcut.pose_estimation_tensorflow.nnet.predict as predict
+        from deeplabcut.pose_estimation_tensorflow.config import load_config
+        from deeplabcut.pose_estimation_tensorflow.nnet import predict_multianimal
+
+        models_folder = 'pose_estimation_tensorflow/models/'
+    # if not DLC 2 is not installed, try import from DLC 1 the old way
+    except ImportError:
+        # adding DLC posing path and loading modules from it
+        sys.path.insert(0,MODEL_PATH + "/pose-tensorflow")
+        from config import load_config
+        from nnet import predict
+        models_folder = 'pose-tensorflow/models/'
 
 
 def load_deeplabcut():
@@ -41,7 +42,7 @@ def load_deeplabcut():
 
     :return: tuple of DeepLabCut config, TensorFlow session, inputs and outputs
     """
-    model = os.path.join(DLC_PATH, models_folder, MODEL)
+    model = os.path.join(MODEL_PATH, models_folder, MODEL_NAME)
     cfg = load_config(os.path.join(model, 'test/pose_cfg.yaml'))
     snapshots = sorted([sn.split('.')[0] for sn in os.listdir(model + '/train/') if "index" in sn])
     cfg['init_weights'] = model + '/train/' + snapshots[-1]
@@ -85,6 +86,7 @@ def find_local_peaks_new(scoremap: np.ndarray, local_reference: np.ndarray, anim
     stride = config['stride']
     # filtering scoremap
     scoremap[scoremap < 0.1] = 0
+
     for joint_num, joint in enumerate(all_joints_names):
         all_peaks[joint] = []
         # selecting the joint in scoremap and locref