Factorize detection merge

examples/anthropogenic-activities/config_trne.yaml

@@ -18,10 +18,10 @@ generate_tilesets.py:
   working_directory: .
   output_folder: output/trne/
   datasets:
-    aoi_tiles: output/trne/tiles.geojson
-    ground_truth_labels: output/trne/labels.geojson
+    aoi_tiles: output/trne/tiles.gpkg
+    ground_truth_labels: output/trne/labels.gpkg
     add_fp_labels:             # Uncomment if FP shapefile exists in prepare_data.py     
-      fp_labels: output/trne/FP.geojson
+      fp_labels: output/trne/FP_labels.gpkg
       frac_trn: 0.7        # fraction of fp tiles to add to the trn dataset, then the remaining tiles will be split in 2 and added to tst and val datasets
     image_source:
       type: XYZ                             # supported values: 1. MIL = Map Image Layer 2. WMS 3. XYZ 4. FOLDER
@@ -82,7 +82,7 @@ assess_detections.py:
   working_directory: output/trne
   output_folder: assessment
   datasets:
-    ground_truth_labels: labels.geojson
+    ground_truth_labels: labels.gpkg
     split_aoi_tiles: split_aoi_tiles.geojson # aoi = Area of Interest
     categories: category_ids.json
     detections:

examples/anthropogenic-activities/merge_detections.py

@@ -9,7 +9,7 @@
 import pandas as pd
 
 sys.path.insert(1, '../..')
-from helpers.functions_for_examples import get_categories
+from helpers.functions_for_examples import get_categories, merge_adjacent_detections, read_dets_and_aoi
 import helpers.misc as misc
 
 from loguru import logger
@@ -62,87 +62,17 @@
         logger.success(f"Done! All files already exist in folder {OUTPUT_DIR}. Exiting.")
         sys.exit(0)
 
-    logger.info("Loading split AoI tiles as a GeoPandas DataFrame...")
-    tiles_gdf = gpd.read_file('split_aoi_tiles.geojson')
-    tiles_gdf = tiles_gdf.to_crs(2056)
-    if 'year_tile' in tiles_gdf.keys(): 
-        tiles_gdf['year_tile'] = tiles_gdf.year_tile.astype(int)
-    logger.success(f"Done! {len(tiles_gdf)} features were found.")
-
-    logger.info("Loading detections as a GeoPandas DataFrame...")
-
-    detections_gdf = gpd.GeoDataFrame()
-
-    for dataset, dets_file in DETECTION_FILES.items():
-        detections_ds_gdf = gpd.read_file(dets_file)    
-        detections_ds_gdf[f'dataset'] = dataset
-        detections_gdf = pd.concat([detections_gdf, detections_ds_gdf], axis=0, ignore_index=True)
-    detections_gdf = detections_gdf.to_crs(2056)
-    detections_gdf['area'] = detections_gdf.area 
-    detections_gdf['det_id'] = detections_gdf.index
-    if 'year_det' in detections_gdf.keys(): 
-        detections_gdf['year_det'] = detections_gdf.year_det.astype(int)
-    logger.success(f"Done! {len(detections_gdf)} features were found.")
+    tiles_gdf, detections_gdf = read_dets_and_aoi(DETECTION_FILES)
 
     # Merge features
     logger.info(f"Merge adjacent polygons overlapping tiles with a buffer of {DISTANCE} m...")
     detections_all_years_gdf = gpd.GeoDataFrame()
 
     # Process detection by year
     for year in detections_gdf.year_det.unique():
-        detections_by_year_gdf = detections_gdf[detections_gdf['year_det']==year]
-
-        detections_buffer_gdf = detections_by_year_gdf.copy()
-        detections_buffer_gdf['geometry'] = detections_by_year_gdf.geometry.buffer(DISTANCE, resolution=2)
-
-        # Saves the id of polygons contained entirely within the tile (no merging with adjacent tiles), to avoid merging them if they are at a distance of less than thd  
-        detections_tiles_join_gdf = gpd.sjoin(tiles_gdf, detections_buffer_gdf, how='left', predicate='contains')
-        remove_det_list = detections_tiles_join_gdf.det_id.unique().tolist()
-
-        detections_within_tiles_gdf = detections_by_year_gdf[detections_by_year_gdf.det_id.isin(remove_det_list)].drop_duplicates(subset=['det_id'], ignore_index=True)
-        detections_overlap_tiles_gdf = detections_by_year_gdf[~detections_by_year_gdf.det_id.isin(remove_det_list)].drop_duplicates(subset=['det_id'], ignore_index=True)
-
-        # Merge polygons within the thd distance
-        detections_overlap_tiles_gdf.loc[:, 'geometry'] = detections_overlap_tiles_gdf.buffer(DISTANCE, resolution=2)
-        detections_dissolve_gdf = detections_overlap_tiles_gdf[['det_id', 'geometry']].dissolve(as_index=False)
-        detections_merge_gdf = detections_dissolve_gdf.explode(ignore_index=True)
-        del detections_dissolve_gdf, detections_overlap_tiles_gdf
-
-        if detections_merge_gdf.isnull().values.any():
-            detections_merge_gdf = gpd.GeoDataFrame()
-        else:
-            detections_merge_gdf.geometry = detections_merge_gdf.buffer(-DISTANCE, resolution=2)
-
-        # Spatially join merged detection with raw ones to retrieve relevant information (score, area,...)
-        detections_merge_gdf['index_merge'] = detections_merge_gdf.index
-        detections_join_gdf = gpd.sjoin(detections_merge_gdf, detections_by_year_gdf, how='inner', predicate='intersects')
-
-        det_class_all = []
-        det_score_all = []
-
-        for id in detections_merge_gdf.index_merge.unique():
-            detections_by_year_gdf = detections_join_gdf.copy()
-            detections_by_year_gdf = detections_by_year_gdf[(detections_by_year_gdf['index_merge']==id)]
-            detections_by_year_gdf.rename(columns={'score_left': 'score'}, inplace=True)
-            det_score_all.append(detections_by_year_gdf['score'].mean())
-            detections_by_year_gdf = detections_by_year_gdf.dissolve(by='det_class', aggfunc='sum', as_index=False)
-            # Keep class of largest det
-            if len(detections_by_year_gdf) > 0:
-                detections_by_year_gdf['det_class'] = detections_by_year_gdf.loc[detections_by_year_gdf['area'] == detections_by_year_gdf['area'].max(), 
-                                                                'det_class'].iloc[0]    
-                det_class = detections_by_year_gdf['det_class'].drop_duplicates().tolist()
-            else:
-                det_class = [0]
-            det_class_all.append(det_class[0])
-
-        detections_merge_gdf['det_class'] = det_class_all
-        detections_merge_gdf['score'] = det_score_all
-
-        complete_merge_dets_gdf = pd.merge(detections_merge_gdf, detections_join_gdf[['index_merge', 'year_det'] + ([] if 'dataset' in detections_merge_gdf.columns else ['dataset'])], on='index_merge')
+        complete_merge_dets_gdf, detections_within_tiles_gdf = merge_adjacent_detections(detections_gdf, tiles_gdf, year, DISTANCE)
         detections_all_years_gdf = pd.concat([detections_all_years_gdf, complete_merge_dets_gdf, detections_within_tiles_gdf], ignore_index=True)
 
-        del complete_merge_dets_gdf, detections_merge_gdf, detections_by_year_gdf, detections_within_tiles_gdf, detections_join_gdf
-
     # get classe ids
     CATEGORIES = os.path.join('category_ids.json')
     categories_info_df, _ = get_categories(CATEGORIES)

examples/anthropogenic-activities/prepare_data.py

@@ -59,32 +59,13 @@
     if not os.path.exists(OUTPUT_DIR):
         os.makedirs(OUTPUT_DIR)
 
-    written_files = []
-
-    gt_labels_4326_gdf = ffe.prepare_labels(SHPFILE, CATEGORY, supercategory=SUPERCATEGORY)
-
-    label_filepath = os.path.join(OUTPUT_DIR, 'labels.geojson')
-    gt_labels_4326_gdf.to_file(label_filepath, driver='GeoJSON')
-    written_files.append(label_filepath)  
-    logger.success(f"Done! A file was written: {label_filepath}")
+    gt_labels_4326_gdf, written_files = ffe.prepare_labels(SHPFILE, CATEGORY, supercategory=SUPERCATEGORY, output_dir=OUTPUT_DIR)
 
-    tiles_4326_all_gdf, tmp_written_files = ffe.format_all_tiles(
-        FP_SHPFILE, os.path.join(OUTPUT_DIR, 'FP.geojson'), EPT_SHPFILE, ept_data_type=EPT, ept_year=EPT_YEAR, labels_4326_gdf=gt_labels_4326_gdf,
-        category=CATEGORY, supercategory=SUPERCATEGORY, zoom_level=ZOOM_LEVEL
+    _, tmp_written_files = ffe.format_all_tiles(
+        FP_SHPFILE, EPT_SHPFILE, ept_data_type=EPT, ept_year=EPT_YEAR, labels_4326_gdf=gt_labels_4326_gdf,
+        category=CATEGORY, supercategory=SUPERCATEGORY, zoom_level=ZOOM_LEVEL, output_dir=OUTPUT_DIR
     )
-
-    # Save tile shapefile
-    tile_filepath = os.path.join(OUTPUT_DIR, 'tiles.geojson')
-    if tiles_4326_all_gdf.empty:
-        logger.warning('No tile generated for the designated area.')
-        tile_filepath = os.path.join(OUTPUT_DIR, 'area_without_tiles.gpkg')
-        gt_labels_4326_gdf.to_file(tile_filepath)
-        written_files.append(tile_filepath)  
-    else:
-        logger.info("Export tiles to GeoJSON (EPSG:4326)...") 
-        tiles_4326_all_gdf.to_file(tile_filepath, driver='GeoJSON')
-        written_files.append(tile_filepath)  
-        logger.success(f"Done! A file was written: {tile_filepath}")
+    written_files.extend(tmp_written_files)
 
     print()
     logger.info("The following files were written. Let's check them out!")

examples/mineral-extract-sites-detection/README.md

@@ -6,13 +6,13 @@ It consists of the following elements:
 - ready-to-use configuration files:
     - `config_trne.yaml`;
     - `config_det.yaml`;
-    - `detectron2_config_dqry.yaml`.
+    - `detectron2_config.yaml`.
 - Input data in the `data` subfolder:
     - MES **labels** issued from the [swissTLM3D](https://www.swisstopo.admin.ch/fr/geodata/landscape/tlm3d.html) product, revised and synchronized with the 2020 [SWISSIMAGE](https://www.swisstopo.admin.ch/fr/geodata/images/ortho/swissimage10.html) orthophotos;
     - the delimitation of the **Area of Interest (AoI)**;
-    - the Swiss DEM raster is too large to be saved on this platform but can be downloaded from this [link](https://github.com/lukasmartinelli/swissdem) using the [EPSG:4326](https://epsg.io/4326) coordinate reference system. The raster must be re-projected to [EPSG:2056](https://epsg.io/2056), renamed as `switzerland_dem_EPSG2056.tif` and located in the **DEM** subfolder. This procedure is managed by running the bash script `get_dem.sh`. 
+    - the Swiss DEM raster can be downloaded from this [link](https://github.com/lukasmartinelli/swissdem) using the [EPSG:4326](https://epsg.io/4326) coordinate reference system. The raster must be re-projected to [EPSG:2056](https://epsg.io/2056), renamed as `switzerland_dem_EPSG2056.tif` and located in the **DEM** subfolder. This procedure is managed by running the bash script `get_dem.sh`. 
 - A data preparation script (`prepare_data.py`) producing the files to be used as input to the `generate_tilesets` stage.
-- A post-processing script (`filter_detections.py`) which filters detections according to their confidence score, altitude and area. The script also identifies and merges groups of nearby polygons.
+- Post-processing scripts (`merge_detections.py` and `filter_detections.py`) which merge adjacent polygons and filter detections according to their confidence score, altitude and area respectively.
 
 The workflow can be run end-to-end by issuing the following list of commands, from the root folder of this GitHub repository:
 
@@ -25,9 +25,11 @@ nobody@<id>:/app# stdl-objdet generate_tilesets config_trne.yaml
 nobody@<id>:/app# stdl-objdet train_model config_trne.yaml
 nobody@<id>:/app# stdl-objdet make_detections config_trne.yaml
 nobody@<id>:/app# stdl-objdet assess_detections config_trne.yaml
+nobody@<id>:/app# python merge_detections.py config_trne.yaml
 nobody@<id>:/app# python prepare_data.py config_det.yaml
 nobody@<id>:/app# stdl-objdet generate_tilesets config_det.yaml
 nobody@<id>:/app# stdl-objdet make_detections config_det.yaml
+nobody@<id>:/app# python merge_detections.py config_det.yaml
 nobody@<id>:/app# bash get_dem.sh
 nobody@<id>:/app# python filter_detections.py config_det.yaml
 nobody@<id>:/app# exit

examples/mineral-extract-sites-detection/config_det.yaml

@@ -1,49 +1,50 @@
-###################################
-####### Inference detection ####### 
-# Automatic detection of Quarries and Mineral Extraction Sites (MES) in images
+##### Detection by inference #####
+# Automatic detection of Mineral Extraction Sites (MES) in aerial images
 
-# 1-Produce tile geometries based on the AoI extent and zoom level
+# Produce tile geometries based on the AoI extent and zoom level
 prepare_data.py: 
   srs: "EPSG:2056"                       # Projection of the input file
   datasets:
     shapefile: ./data/AoI/AoI_2020.shp
   output_folder: ./output/det/
   zoom_level: 16   
 
-# 2-Fetch of tiles (online server) and split into 3 datasets: train, test, validation
+# Fetch of tiles (online server) and split into 3 datasets: train, test, validation
 generate_tilesets.py:
   debug_mode: 
     enable: False  # sample of tiles
     nb_tiles_max: 5000
-  working_directory: output
+  working_directory: ./output/
   datasets:
-    aoi_tiles: det/tiles.geojson
+    aoi_tiles: det/tiles.gpkg
     image_source:
-      type: XYZ     # supported values: 1. MIL = Map Image Layer 2. WMS 3. XYZ 4. FOLDER
-      location: https://wmts.geo.admin.ch/1.0.0/ch.swisstopo.swissimage-product/default/2020/3857/{z}/{x}/{y}.jpeg
+      type: XYZ                             # supported values: 1. MIL = Map Image Layer 2. WMS 3. XYZ 4. FOLDER
+      year: 2020                            # supported values: 1. multi-year (tiles of different year), 2. <year> (i.e. 2020)
+      location: https://wmts.geo.admin.ch/1.0.0/ch.swisstopo.swissimage-product/default/{year}/3857/{z}/{x}/{y}.jpeg
   output_folder: det/
   tile_size: 256    # per side, in pixels
   overwrite: False
   n_jobs: 10
+  seed: 42
   COCO_metadata:
     year: 2021
     version: 1.0
-    description: Swiss Image Hinterground w/ Quarries and Mineral Exploitation Sites detection
+    description: Swissimage background with Mineral Exploitation Site labels
     contributor: swisstopo
     url: https://swisstopo.ch
     license:
-      name: Unknown
-      url:
+      name: unknown
+      url: unknown
     categories_file: trne/category_ids.json
 
-# 3-Object detection by inference with the optimised trained model
+# Object detection by inference with the optimised trained model
 make_detections.py:
   working_directory: ./output/det/
   log_subfolder: logs
   sample_tagged_img_subfolder: sample_tagged_images
-  COCO_files:           # relative paths, w/ respect to the working_folder
+  COCO_files:           # relative paths, with respect to the working_folder
     oth: COCO_oth.json
-  detectron2_config_file: ../../detectron2_config_dqry.yaml # path relative to the working_folder
+  detectron2_config_file: ../../detectron2_config.yaml # path relative to the working_folder
   model_weights:
     pth_file: ../trne/logs/model_final.pth # trained model minimising the validation loss curve, monitor the training process via tensorboard (tensorboard --logdir </logs>) 
   image_metadata_json: img_metadata.json
@@ -53,14 +54,21 @@ make_detections.py:
   score_lower_threshold: 0.3
   remove_det_overlap: False  # if several detections overlap (IoU > 0.5), only the one with the highest confidence score is retained. Not recommended for use with a single class.
 
-# 4-Filtering and merging detection polygons
+# Assess the final results
+merge_detections.py:
+  working_directory: ./output/det/
+  detections:
+    oth: oth_detections_at_0dot3_threshold.gpkg
+  distance: 10 # m, distance use as a buffer to merge close polygons (likely to belong to the same object) together
+  iou_threshold: 0.1
+  score_threshold: 0.3 # choose a value
+
+# Filtering and merging detection polygons
 filter_detections.py:
-  year: 2020
-  detections: ./output/det/oth_detections_at_0dot3_threshold.gpkg
-  shapefile: ./data/AoI/AoI_2020.shp
-  dem: ./data/DEM/switzerland_dem_EPSG2056.tif
-  elevation: 1200.0   # m, altitude threshold
-  score: 0.95         # detection score (from 0 to 1) provided by detectron2
-  distance: 10        # m, distance use as a buffer to merge close polygons (likely to belong to the same object) together
-  area: 5000.0        # m2, area threshold under which polygons are discarded
-  output: ./output/det/oth_detections_at_0dot3_threshold_year-{year}_score-{score}_area-{area}_elevation-{elevation}_distance-{distance}.geojson
+  working_directory: ./data
+  detections: ../output/det/merged_detections_at_0dot3_threshold.gpkg
+  aoi: ./AoI/AoI_2020.shp
+  dem: ./DEM/switzerland_dem_EPSG2056.tif
+  elevation_threshold: 1200 # m, altitude threshold, optional 
+  score_threshold: 0.95 # detection score (from 0 to 1) provided by detectron2, optional
+  area_threshold: 5000 # m2, area threshold under which polygons are discarded, optional