added OUTPUT_PATH

annotation_tab/annotation_logic.py

@@ -2,11 +2,11 @@ import os
 import pandas as pd
 import threading
 import gradio as gr
-
+from config import OUTPUT_DIR
 
 # ==== CONFIG ====
-IMAGE_FOLDER = "output/blobs"
-CSV_FILE = "output/manual_annotations.csv"
+IMAGE_FOLDER = os.path.join(OUTPUT_DIR,"blobs")
+CSV_FILE = os.path.join(OUTPUT_DIR,"annotations")
 
 # ==== STATE ====
 if os.path.exists(CSV_FILE):

config.py

@@ -0,0 +1,5 @@
+import os
+
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+OUTPUT_DIR = os.path.join(BASE_DIR, "output")
+os.makedirs(OUTPUT_DIR, exist_ok=True)

inference_tab/inference_logic.py

@@ -16,7 +16,7 @@ from osgeo import gdal
 import geopandas as gpd
 from rapidfuzz import process, fuzz
 from huggingface_hub import hf_hub_download
-
+from config import OUTPUT_DIR
 
 yolo_weights = hf_hub_download(
     repo_id="muk42/yolov9_streets",
@@ -33,7 +33,8 @@ def run_inference(image_path, gcp_path, city_name, score_th):
     yield from blobsOCR(image_path)
     
     # === ADD GEO DATA ===
-    yield from georefImg("output/mask.tif", gcp_path)
+    MASK_PATH = os.path.join(OUTPUT_DIR,"mask.tif")
+    yield from georefImg(MASK_PATH, gcp_path)
     yield from extractCentroids(image_path)
     yield from extractStreetNet(city_name)
     
@@ -106,12 +107,12 @@ def getBBoxes(image_path, tile_size=256, overlap=0.3, confidence_threshold=0.25)
     ]
 
     
-    output_path = f"output/boxes.json"
-    os.makedirs("output", exist_ok=True)
-    with open(output_path, "w") as f:
+    BOXES_PATH = os.path.join(OUTPUT_DIR,"boxes.json")
+    
+    with open(BOXES_PATH, "w") as f:
         json.dump(boxes_to_save, f, indent=4)
 
-    yield f"Inference complete. Results saved to {output_path}"
+    yield f"Inference complete."
 
 
 def box_inside_global(box, global_box):
@@ -325,7 +326,8 @@ def getSegments(image_path,iou=0.5,c_th=0.75,edge_margin=10):
     model = SAM("sam2.1_l.pt")
 
     # Load YOLO-predicted boxes
-    with open(f"output/boxes.json", "r") as f:
+    BOXES_PATH = os.path.join(OUTPUT_DIR,"boxes.json")
+    with open(BOXES_PATH, "r") as f:
         box_data = json.load(f)
 
 
@@ -474,7 +476,8 @@ def getSegments(image_path,iou=0.5,c_th=0.75,edge_margin=10):
             
     
     final_mask = Image.fromarray(full_mask)
-    final_mask.save(f"output/mask.tif")
+    MASK_PATH = os.path.join(OUTPUT_DIR,"mask.tif")
+    final_mask.save(MASK_PATH)
 
     yield f"Saved mask with {instance_id - 1} instances"
     
@@ -483,7 +486,8 @@ def getSegments(image_path,iou=0.5,c_th=0.75,edge_margin=10):
 def extractSegments(image_path, min_size=500, margin=10):
    
     image = cv2.imread(image_path)
-    mask = cv2.imread(f"output/mask.tif", cv2.IMREAD_UNCHANGED) 
+    MASK_PATH = os.path.join(OUTPUT_DIR,"mask.tif")
+    mask = cv2.imread(MASK_PATH, cv2.IMREAD_UNCHANGED) 
    
     height, width = mask.shape[:2]
 
@@ -524,9 +528,8 @@ def extractSegments(image_path, min_size=500, margin=10):
             masked_image = cv2.bitwise_and(cropped_image, cropped_image, mask=cropped_mask)
 
         # Save the masked image
-        output_path = os.path.join('output/blobs', f"{blob_id}.png")
-        os.makedirs(os.path.dirname(output_path), exist_ok=True)
-        cv2.imwrite(output_path, masked_image)
+        BLOB_PATH=os.path.join(OUTPUT_DIR,"blobs",f"{blob_id}.png")
+        cv2.imwrite(BLOB_PATH, masked_image)
         
     yield f"Done."
 
@@ -548,9 +551,10 @@ def blobsOCR(image_path):
 
 
     # Open output file for writing
-    with open(f"output/ocr", "w", encoding="utf-8") as f_out:
+    OCR_PATH = os.path.join(OUTPUT_DIR,"ocr")
+    with open(OCR_PATH, "w", encoding="utf-8") as f_out:
         # Process each image
-        image_folder = "output/blobs"
+        image_folder = os.path.join(OUTPUT_DIR,"blobs")
         for filename in os.listdir(image_folder):
             if filename.lower().endswith(image_extensions):
                 image_path = os.path.join(image_folder, filename)
@@ -575,8 +579,8 @@ def blobsOCR(image_path):
 
 
 def extractCentroids(image_path):
-    
-    with rasterio.open(f"output/georeferenced.tif") as src:
+    GEO_PATH=os.path.join(OUTPUT_DIR,"georeferenced.tif")
+    with rasterio.open(GEO_PATH) as src:
         mask = src.read(1)
         transform = src.transform
 
@@ -612,22 +616,13 @@ def extractCentroids(image_path):
         data.append({"blob_id": label, "x": centroid.x, "y": centroid.y})
 
     df = pd.DataFrame(data)
-    df.to_csv(f"output/centroids.csv", index=False)
+    COORD_PATH=os.path.join(OUTPUT_DIR,"centroids.csv")
+    df.to_csv(COORD_PATH, index=False)
     yield f"Saved centroid coordinates of {len(labels)} blobs."
 
 
 
-def collectBlobs(image_path):
-    filename = os.path.splitext(os.path.basename(image_path))[0]
-    box_dir = "output/blobs"
-    # Get all filenames in the folder (only files, not subfolders)
-    file_names = [f for f in os.listdir(box_dir) if os.path.isfile(os.path.join(box_dir, f))]
 
-    # Save to text file
-    with open(f"output/{filename}_blobs.txt", "w") as f:
-        for name in file_names:
-            yield f"Writing {name}..."
-            f.write(name + "\n")
 
 def img_shape(image_path):
     img = cv2.imread(image_path)
@@ -656,7 +651,7 @@ def georefImg(image_path, gcp_path):
 
     
    
-    tmp_file = "output/tmp.tif"
+    tmp_file = os.path.join(OUTPUT_DIR,"tmp.tif")
     
     gdal.Translate(
             tmp_file,
@@ -668,7 +663,7 @@ def georefImg(image_path, gcp_path):
 
    
 
-    geo_file = "output/georeferenced.tif"
+    geo_file = os.path.join(OUTPUT_DIR,"georeferenced.tif")
     yield "Running gdalwarp..."
 
     gdal.Warp(
@@ -691,7 +686,8 @@ def extractStreetNet(city_name):
     G_proj = ox.project_graph(G)
     nodes, edges = ox.graph_to_gdfs(G_proj)
     edges_3857 = edges.to_crs(epsg=3857)
-    edges_3857.to_file("output/osm_extract.geojson", driver="GeoJSON")
+    OSM_PATH=os.path.join(OUTPUT_DIR,"osm_extract.geojson")
+    edges_3857.to_file(OSM_PATH, driver="GeoJSON")
     yield "Done."
 
 
@@ -707,8 +703,10 @@ def best_street_match(point, query_name, edges_gdf, max_distance=100):
         return best_match  # (name, score, index)
 
 def fuzzyMatch():
-    coords_df = pd.read_csv("output/centroids.csv")
-    names_df = pd.read_csv("output/ocr.csv",sep="\t",columns=[['blob_id','pred_text']]) 
+    COORD_PATH=os.path.join(OUTPUT_DIR,"centroids.csv")
+    OCR_PATH=os.path.join(OUTPUT_DIR,"ocr.csv")
+    coords_df = pd.read_csv(COORD_PATH)
+    names_df = pd.read_csv(OCR_PATH,sep="\t",columns=[['blob_id','pred_text']]) 
     merged_df = coords_df.merge(names_df, on="blob_id")
 
     gdf = gpd.GeoDataFrame(
@@ -717,7 +715,8 @@ def fuzzyMatch():
         crs="EPSG:3857"
     )
 
-    osm_gdf = gpd.read_file("output/osm_extract.geojson")
+    OSM_PATH=os.path.join(OUTPUT_DIR,"osm_extract.geojson")
+    osm_gdf = gpd.read_file(OSM_PATH)
     osm_gdf = osm_gdf[osm_gdf['name'].notnull()]
 
     yield "Process OSM candidates..."
@@ -744,5 +743,6 @@ def fuzzyMatch():
             })
 
     results_df = pd.DataFrame(results)
-    results_df.to_csv("output/street_matches.csv", index=False)
-    yield "output/street_matches.csv"
+    RES_PATH=os.path.join(OUTPUT_DIR,"street_matches.csv")
+    results_df.to_csv(RES_PATH, index=False)
+    yield f"{RES_PATH}/street_matches.csv"