Make import of external SWIN and EmoNet packages optional

gdl/layers/losses/EmoNetLoss.py

@@ -5,8 +5,14 @@
 from gdl.layers.losses.EmonetLoader import get_emonet
 from pathlib import Path
 import torch.nn.functional as F
-from gdl.models.EmoNetModule import EmoNetModule
-from gdl.models.EmoSwinModule import EmoSwinModule
+try:
+    from gdl.models.EmoNetModule import EmoNetModule
+except ImportError as e:
+    print(f"Could not import EmoNetModule. EmoNet models will not be available.")
+try:
+    from gdl.models.EmoSwinModule import EmoSwinModule
+except ImportError as e: 
+    print(f"Could not import EmoSwinModule. SWIN models will not be available")
 from gdl.models.EmoCnnModule import EmoCnnModule
 from gdl.models.IO import get_checkpoint_with_kwargs
 from gdl.utils.other import class_from_str

gdl/models/DecaEncoder.py

@@ -6,7 +6,10 @@
 import torch.nn.functional as F
 import gdl.models.ResNet as resnet
 
-from .Swin import create_swin_backbone, swin_cfg_from_name
+try:
+    from .Swin import create_swin_backbone, swin_cfg_from_name
+except ImportError as e:
+    print("SWIN not found, will not be able to use SWIN models")
 
 class BaseEncoder(nn.Module):
     def __init__(self, outsize, last_op=None):

gdl_apps/EMOCA/demos/test_emoca_on_video.py

@@ -59,29 +59,29 @@ def main():
     ## 3) Get the data loadeer with the detected faces
     dl = dm.test_dataloader()
 
-    # ## 4) Run the model on the data
-    # for j, batch in enumerate (auto.tqdm( dl)):
+    ## 4) Run the model on the data
+    for j, batch in enumerate (auto.tqdm( dl)):
 
-    #     current_bs = batch["image"].shape[0]
-    #     img = batch
-    #     vals, visdict = test(emoca, img)
-    #     for i in range(current_bs):
-    #         # name = f"{(j*batch_size + i):05d}"
-    #         name =  batch["image_name"][i]
+        current_bs = batch["image"].shape[0]
+        img = batch
+        vals, visdict = test(emoca, img)
+        for i in range(current_bs):
+            # name = f"{(j*batch_size + i):05d}"
+            name =  batch["image_name"][i]
 
-    #         sample_output_folder = Path(outfolder) /name
-    #         sample_output_folder.mkdir(parents=True, exist_ok=True)
+            sample_output_folder = Path(outfolder) /name
+            sample_output_folder.mkdir(parents=True, exist_ok=True)
 
-    #         if args.save_mesh:
-    #             save_obj(emoca, str(sample_output_folder / "mesh_coarse.obj"), vals, i)
-    #         if args.save_images:
-    #             save_images(outfolder, name, visdict, i)
-    #         if args.save_codes:
-    #             save_codes(Path(outfolder), name, vals, i)
+            if args.save_mesh:
+                save_obj(emoca, str(sample_output_folder / "mesh_coarse.obj"), vals, i)
+            if args.save_images:
+                save_images(outfolder, name, visdict, i)
+            if args.save_codes:
+                save_codes(Path(outfolder), name, vals, i)
 
-    # ## 5) Create the reconstruction video (reconstructions overlayed on the original video)
-    # dm.create_reconstruction_video(0,  rec_method=model_name, image_type=image_type, overwrite=True)
-    # print("Done")
+    ## 5) Create the reconstruction video (reconstructions overlayed on the original video)
+    dm.create_reconstruction_video(0,  rec_method=model_name, image_type=image_type, overwrite=True)
+    print("Done")
 
 
 if __name__ == '__main__':

gdl_apps/EMOCA/deprecated/expression_disentanglement_experiment.py

@@ -7,7 +7,10 @@
 
 from gdl.models.DECA import DECA, ExpDECA, DecaModule
 from gdl.models.IO import locate_checkpoint
-from gdl.models.EmoNetModule import EmoNetModule
+try:
+    from gdl.models.EmoNetModule import EmoNetModule
+except ImportError as e: 
+    print("Skipping EmoNetModule because EmoNet it is not installed")
 from gdl.utils.other import class_from_str
 import datetime
 from pytorch_lightning import Trainer

gdl_apps/EmotionRecognition/evaluation/compute_confusion_matrix.py

@@ -7,7 +7,10 @@
 from gdl.models.IO import locate_checkpoint, get_checkpoint_with_kwargs
 
 from gdl.models.EmoDECA import EmoDECA
-from gdl.models.EmoNetModule import EmoNetModule
+try:
+    from gdl.models.EmoNetModule import EmoNetModule
+except ImportError as e: 
+    print("Skipping EmoNetModule because EmoNet it is not installed")
 from gdl.utils.other import class_from_str
 import datetime
 from pytorch_lightning import Trainer

gdl_apps/EmotionRecognition/training/train_emodeca.py

@@ -16,9 +16,15 @@
 from gdl.models.IO import locate_checkpoint, get_checkpoint_with_kwargs
 
 from gdl.models.EmoDECA import EmoDECA
-from gdl.models.EmoSwinModule import EmoSwinModule
+try:
+    from gdl.models.EmoSwinModule import EmoSwinModule
+except ImportError as e: 
+    print(f"Could not import EmoSwinModule. SWIN models will not be available")
 from gdl.models.EmoCnnModule import EmoCnnModule
-from gdl.models.EmoNetModule import EmoNetModule
+try:
+    from gdl.models.EmoNetModule import EmoNetModule
+except ImportError as e: 
+    print(f"Could not import EmoNet. EmoNet models will not be available")
 from gdl.models.EmoMLP import EmoMLP
 
 from gdl.utils.other import class_from_str

gdl_apps/EmotionRecognition/utils/io.py

@@ -1,7 +1,13 @@
 from gdl.models.EmoDECA import EmoDECA 
 from gdl.models.EmoCnnModule import EmoCnnModule 
-from gdl.models.EmoSwinModule import EmoSwinModule 
-from gdl.models.EmoNetModule import EmoNetModule
+try:
+    from gdl.models.EmoSwinModule import EmoSwinModule 
+except ImportError as e: 
+    print(f"Could not import EmoSwinModule. SWIN models will not be available")
+try:
+    from gdl.models.EmoNetModule import EmoNetModule
+except ImportError as e: 
+    print(f"Could not import EmoNetModule. EmoNet models will not be available")
 from gdl.models.IO import locate_checkpoint
 from gdl.utils.other import class_from_str
 from pathlib import Path