Fixx ida lstm in taasss dataset

checkpoints

+/data1/IDA_LSTM_checkpoints
\ No newline at end of file

cikm_inter_dst_predrnn_run_taasss.py

@@ -6,10 +6,12 @@ import numpy as np
 from tqdm import tqdm
 
 import core.trainer as trainer
-from cikm_inter_dst_predrnn_run_taasss_utils import get_batcher, padding_taasss
+from cikm_inter_dst_predrnn_run_taasss_utils import get_batcher, padding_taasss, change_taasss_dims
 from core.models.model_factory import Model
 from core.utils import preprocess
-
+from matplotlib import pyplot as plt
+import scipy.misc
+from PIL import Image
 
 parser = argparse.ArgumentParser(
     description="PyTorch video prediction model - DST PredRNN"
@@ -126,14 +128,6 @@ def schedule_sampling(eta: float, itr: int):
     )
     return eta, real_input_flag
 
-
-def change_taasss_dims(a: np.ndarray) -> np.ndarray:
-    """(25, 1, 1, 480, 480) to (1, 25, 480, 480, 1)"""
-    a = np.squeeze(a)
-    a = np.expand_dims(a, axis=0)
-    return np.expand_dims(a, axis=4)
-
-
 def wrapper_train(model: Model):
     eta = args.sampling_start_value
     iterator = get_batcher(args)
@@ -143,8 +137,6 @@ def wrapper_train(model: Model):
         imgs = change_taasss_dims(imgs)
         imgs = padding_taasss(imgs, args)
         imgs = preprocess.reshape_patch(imgs, args.patch_size)
-        # Should already by 0 to 1
-        # imgs = nor(imgs)
         eta, real_input_flag = schedule_sampling(eta, itr)
         cost = trainer.train(model, imgs, real_input_flag, args)
         progress_bar.set_description(f"Loss: {cost}")

cikm_inter_dst_predrnn_run_taasss_utils.py

@@ -14,7 +14,9 @@ def get_batcher(args):
         "r",
         libver="latest",
     )
-    outlier_mask = cv2.imread(str(data_dir / "mask.png"), 0)
+    mask_path = data_dir / "mask.png"
+    assert mask_path.exists(), "Mask does not exist"
+    outlier_mask = cv2.imread(str(mask_path), 0)
     metadata = pd.read_csv(metadata_file, index_col="id")
     metadata["start_datetime"] = pd.to_datetime(metadata["start_datetime"])
     metadata["end_datetime"] = pd.to_datetime(metadata["end_datetime"])
@@ -30,7 +32,7 @@ def get_batcher(args):
     )
     # Filter images in the ROI
     for imgs, _, masks in batcher:
-        yield imgs * ~masks
+        yield imgs * masks
 
 
 def padding_taasss(array: np.ndarray, args) -> np.ndarray:
@@ -64,3 +66,8 @@ def unpadding_taasss(array: np.ndarray, args) -> np.ndarray:
     elif args.img_width == 128:
         return array[:, :, 4:124, 4:124, :]
     raise ValueError
+def change_taasss_dims(a: np.ndarray) -> np.ndarray:
+    """(25, 1, 1, 480, 480) to (1, 25, 480, 480, 1)"""
+    a = np.squeeze(a)
+    a = np.expand_dims(a, axis=0)
+    return np.expand_dims(a, axis=4)

requirements.txt

@@ -6,4 +6,5 @@ imageio
 pandas
 h5py
 tqdm
-matplotlib
\ No newline at end of file
+matplotlib
+pysteps
\ No newline at end of file