feat:data-custom-one-1

.gitignore

@@ -7,7 +7,11 @@ __pycache__/
 
 # C extensions
 *.so
-
+/custom_data_set/*.npy
+/custom_data_set/*.txt
+/hand_gen_37_torch/
+/data/*.json
+/logs/
 # Distribution / packaging
 .Python
 build/

app/priming.py

@@ -27,8 +27,8 @@ def generate_handwriting(
     bias=10.0,
 ):
     device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
-    data_path = os.path.join(app_path, "../data/")
-    model_path = os.path.join(app_path, "../results/synthesis/best_model_synthesis.pt")
+    data_path = os.path.join(app_path, "../custom_data_set/")
+    model_path = os.path.join(app_path, "/media/hassan/New Volume1/Remote_work_Research_work/Handwriting-synthesis/logs/best_model_own_datasynthesis.pt")
     # seed = 194
     # print("seed:",seed)
     # torch.manual_seed(seed)

app/static/uploads/33b7d4a9-2246-48ed-a68c-727f75841ab2/inpText.txt

@@ -0,0 +1 @@
+hello world

custom_data_set/prepare_data_json.py

@@ -0,0 +1,83 @@
+import numpy as np
+import json
+
+# Load your JSON data
+def convert_json_to_repo_format(json_file_path, output_npy_path, output_txt_path):
+    # Load JSON data
+    with open(json_file_path, 'r') as f:
+        data = json.load(f)
+
+    # Initialize lists to store processed data
+    all_strokes = []
+    all_prompts = []
+
+    # Calculate mean and std for normalization
+    all_x = []
+    all_y = []
+
+    # First pass to collect all x, y coordinates for normalization
+    for entry in data:
+        stroke_data = entry['stroke_data']
+        for stroke in stroke_data:
+            for point in stroke:
+                all_x.append(point['x'])
+                all_y.append(point['y'])
+
+    # Calculate mean and std
+    x_mean = np.mean(all_x)
+    y_mean = np.mean(all_y)
+    x_std = np.std(all_x)
+    y_std = np.std(all_y)
+
+    print(f"X mean: {x_mean}, X std: {x_std}")
+    print(f"Y mean: {y_mean}, Y std: {y_std}")
+
+    # Second pass to process the data
+    for entry in data:
+        prompt_text = entry['prompt_text']
+        stroke_data = entry['stroke_data']
+
+        # Convert stroke data to the required format
+        formatted_strokes = []
+
+        for stroke in stroke_data:
+            for i, point in enumerate(stroke):
+                # Normalize x and y
+                x_norm = (point['x'] - x_mean) / x_std
+                y_norm = (point['y'] - y_mean) / y_std
+
+                # Last point in a stroke gets 1, others get 0
+                end_of_stroke = 1 if i == len(stroke) - 1 else 0
+
+                formatted_strokes.append([end_of_stroke, x_norm, y_norm])
+
+        # Convert to numpy array
+        formatted_strokes = np.array(formatted_strokes)
+
+        all_strokes.append(formatted_strokes)
+        all_prompts.append(prompt_text)
+
+    # Save the processed data
+    # Convert all_strokes to a numpy array with dtype=object since each stroke array has different shape
+    all_strokes_array = np.array(all_strokes, dtype=object)
+    np.save(output_npy_path, all_strokes_array)
+
+    # Write prompts to text file
+    with open(output_txt_path, 'w') as f:
+        f.write('\n'.join(all_prompts))
+
+    print(f"Saved {len(all_strokes)} entries to {output_npy_path} and {output_txt_path}")
+
+    # Print sample information
+    if len(all_strokes) > 0:
+        print(f"Shape of the first entry: {all_strokes[0].shape}")
+        print(f"Sample data from first entry:")
+        print(all_strokes[0][:3])  # First 3 points
+        print(all_strokes[0][-3:])  # Last 3 points
+
+# Usage
+convert_json_to_repo_format(
+    '/media/hassan/New Volume1/Remote_work_Research_work/Handwriting-synthesis/data/hand_1000_data.json',
+    'strokes_data.npy',
+    'sentences_data.txt'
+)

train.py

@@ -36,7 +36,7 @@ def argparser():
     parser.add_argument("--lr", type=float, default=0.001)
     parser.add_argument("--patience", type=int, default=15)
     parser.add_argument("--model_type", type=str, default="prediction")
-    parser.add_argument("--data_path", type=str, default="./data/")
+    parser.add_argument("--data_path", type=str, default="./custom_data_set/")
     parser.add_argument("--save_path", type=str, default="./logs/")
     parser.add_argument("--text_req", action="store_true")
     parser.add_argument("--data_aug", action="store_true")
@@ -163,7 +163,7 @@ def train(
     model_type,
     save_path,
 ):
-    model_path = save_path + "best_model_" + model_type + ".pt"
+    model_path = save_path + "best_model_own_data" + model_type + ".pt"
     model = model.to(device)
 
     optimizer = optim.Adam(model.parameters(), lr=lr)