MNT use np.clip in motion energy extraction

Author: TomDLT

tutorials/movies_3T/06_extract_motion_energy.py

@@ -135,8 +135,7 @@ def compute_motion_energy(luminance,
         # add some noise to deal with constant black areas
         luminance_batch = luminance[batch].copy()
         luminance_batch += np.random.randn(*luminance_batch.shape) * noise
-        luminance_batch[luminance_batch < 0] = 0
-        luminance_batch[luminance_batch > 100] = 100
+        luminance_batch = np.clip(luminance_batch, 0, 100)
 
         motion_energy[batch] = pyramid.project_stimulus(luminance_batch)
 

tutorials/movies_4T/01_extract_motion_energy.py

@@ -135,8 +135,7 @@ def compute_motion_energy(luminance,
         # add some noise to deal with constant black areas
         luminance_batch = luminance[batch].copy()
         luminance_batch += np.random.randn(*luminance_batch.shape) * noise
-        luminance_batch[luminance_batch < 0] = 0
-        luminance_batch[luminance_batch > 100] = 100
+        luminance_batch = np.clip(luminance_batch, 0, 100)
 
         motion_energy[batch] = pyramid.project_stimulus(luminance_batch)
 

tutorials/notebooks/movies_3T/06_extract_motion_energy.ipynb

@@ -91,7 +91,7 @@
       },
       "outputs": [],
       "source": [
-        "from scipy.signal import decimate\nfrom moten.pyramids import MotionEnergyPyramid\n\n# fixed experiment settings\nN_FRAMES_PER_SEC = 15\nN_FRAMES_PER_TR = 30\nN_TRS_PER_RUN = 300\n\n\ndef compute_motion_energy(luminance,\n                          batch_size=N_TRS_PER_RUN * N_FRAMES_PER_TR,\n                          noise=0.1):\n\n    n_frames, height, width = luminance.shape\n\n    # We create a pyramid instance, with the main motion-energy parameters.\n    pyramid = MotionEnergyPyramid(stimulus_vhsize=(height, width),\n                                  stimulus_fps=N_FRAMES_PER_SEC,\n                                  spatial_frequencies=[0, 2, 4, 8, 16, 32])\n\n    # We batch images run by run.\n    motion_energy = np.zeros((n_frames, pyramid.nfilters))\n    for ii, start in enumerate(range(0, n_frames, batch_size)):\n        batch = slice(start, start + batch_size)\n        print(\"run %d\" % ii)\n\n        # add some noise to deal with constant black areas\n        luminance_batch = luminance[batch].copy()\n        luminance_batch += np.random.randn(*luminance_batch.shape) * noise\n        luminance_batch[luminance_batch < 0] = 0\n        luminance_batch[luminance_batch > 100] = 100\n\n        motion_energy[batch] = pyramid.project_stimulus(luminance_batch)\n\n    # decimate to the sampling frequency of fMRI responses\n    motion_energy_decimated = decimate(motion_energy, N_FRAMES_PER_TR,\n                                       ftype='fir', axis=0)\n    return motion_energy_decimated\n\n\nmotion_energy_train = compute_motion_energy(luminance_train)\nmotion_energy_test = compute_motion_energy(luminance_test)"
+        "from scipy.signal import decimate\nfrom moten.pyramids import MotionEnergyPyramid\n\n# fixed experiment settings\nN_FRAMES_PER_SEC = 15\nN_FRAMES_PER_TR = 30\nN_TRS_PER_RUN = 300\n\n\ndef compute_motion_energy(luminance,\n                          batch_size=N_TRS_PER_RUN * N_FRAMES_PER_TR,\n                          noise=0.1):\n\n    n_frames, height, width = luminance.shape\n\n    # We create a pyramid instance, with the main motion-energy parameters.\n    pyramid = MotionEnergyPyramid(stimulus_vhsize=(height, width),\n                                  stimulus_fps=N_FRAMES_PER_SEC,\n                                  spatial_frequencies=[0, 2, 4, 8, 16, 32])\n\n    # We batch images run by run.\n    motion_energy = np.zeros((n_frames, pyramid.nfilters))\n    for ii, start in enumerate(range(0, n_frames, batch_size)):\n        batch = slice(start, start + batch_size)\n        print(\"run %d\" % ii)\n\n        # add some noise to deal with constant black areas\n        luminance_batch = luminance[batch].copy()\n        luminance_batch += np.random.randn(*luminance_batch.shape) * noise\n        luminance_batch = np.clip(luminance_batch, 0, 100)\n\n        motion_energy[batch] = pyramid.project_stimulus(luminance_batch)\n\n    # decimate to the sampling frequency of fMRI responses\n    motion_energy_decimated = decimate(motion_energy, N_FRAMES_PER_TR,\n                                       ftype='fir', axis=0)\n    return motion_energy_decimated\n\n\nmotion_energy_train = compute_motion_energy(luminance_train)\nmotion_energy_test = compute_motion_energy(luminance_test)"
       ]
     },
     {

tutorials/notebooks/movies_4T/01_extract_motion_energy.ipynb

@@ -98,7 +98,7 @@
       },
       "outputs": [],
       "source": [
-        "from scipy.signal import decimate\nfrom moten.pyramids import MotionEnergyPyramid\n\n# fixed experiment settings\nN_FRAMES_PER_SEC = 15\nN_FRAMES_PER_TR = 15\nN_TRS_PER_RUN = 600\n\n\ndef compute_motion_energy(luminance,\n                          batch_size=N_TRS_PER_RUN * N_FRAMES_PER_TR,\n                          noise=0.1):\n\n    n_frames, height, width = luminance.shape\n\n    # We create a pyramid instance, with the main motion-energy parameters.\n    pyramid = MotionEnergyPyramid(stimulus_vhsize=(height, width),\n                                  stimulus_fps=N_FRAMES_PER_SEC,\n                                  spatial_frequencies=[0, 2, 4, 8, 16, 32])\n\n    # We batch images run by run.\n    motion_energy = np.zeros((n_frames, pyramid.nfilters))\n    for ii, start in enumerate(range(0, n_frames, batch_size)):\n        batch = slice(start, start + batch_size)\n        print(\"run %d\" % ii)\n\n        # add some noise to deal with constant black areas\n        luminance_batch = luminance[batch].copy()\n        luminance_batch += np.random.randn(*luminance_batch.shape) * noise\n        luminance_batch[luminance_batch < 0] = 0\n        luminance_batch[luminance_batch > 100] = 100\n\n        motion_energy[batch] = pyramid.project_stimulus(luminance_batch)\n\n    # decimate to the sampling frequency of fMRI responses\n    motion_energy_decimated = decimate(motion_energy, N_FRAMES_PER_TR,\n                                       ftype='fir', axis=0)\n    return motion_energy_decimated\n\n\nmotion_energy_train = compute_motion_energy(luminance_train)\nmotion_energy_test = compute_motion_energy(luminance_test)"
+        "from scipy.signal import decimate\nfrom moten.pyramids import MotionEnergyPyramid\n\n# fixed experiment settings\nN_FRAMES_PER_SEC = 15\nN_FRAMES_PER_TR = 15\nN_TRS_PER_RUN = 600\n\n\ndef compute_motion_energy(luminance,\n                          batch_size=N_TRS_PER_RUN * N_FRAMES_PER_TR,\n                          noise=0.1):\n\n    n_frames, height, width = luminance.shape\n\n    # We create a pyramid instance, with the main motion-energy parameters.\n    pyramid = MotionEnergyPyramid(stimulus_vhsize=(height, width),\n                                  stimulus_fps=N_FRAMES_PER_SEC,\n                                  spatial_frequencies=[0, 2, 4, 8, 16, 32])\n\n    # We batch images run by run.\n    motion_energy = np.zeros((n_frames, pyramid.nfilters))\n    for ii, start in enumerate(range(0, n_frames, batch_size)):\n        batch = slice(start, start + batch_size)\n        print(\"run %d\" % ii)\n\n        # add some noise to deal with constant black areas\n        luminance_batch = luminance[batch].copy()\n        luminance_batch += np.random.randn(*luminance_batch.shape) * noise\n        luminance_batch = np.clip(luminance_batch, 0, 100)\n\n        motion_energy[batch] = pyramid.project_stimulus(luminance_batch)\n\n    # decimate to the sampling frequency of fMRI responses\n    motion_energy_decimated = decimate(motion_energy, N_FRAMES_PER_TR,\n                                       ftype='fir', axis=0)\n    return motion_energy_decimated\n\n\nmotion_energy_train = compute_motion_energy(luminance_train)\nmotion_energy_test = compute_motion_energy(luminance_test)"
       ]
     },
     {