Use deepforest config file and deepforest's boxes_to_shapefile

deepforest_config.yml

@@ -0,0 +1,40 @@
+# Config file for DeepForest pytorch module
+
+# Cpu workers for data loaders
+# Dataloaders
+workers: 1
+devices: 1
+accelerator: 'gpu'
+batch_size: 1
+
+# Model Architecture
+architecture: 'retinanet'
+num_classes: 1
+nms_thresh: 0.05
+
+# Architecture specific params
+retinanet:
+    # Non-max supression of overlapping predictions
+    score_thresh: 0.1
+
+train:
+    csv_file:
+    root_dir:
+
+    # Optimizer initial learning rate
+    lr: 0.001
+
+    # Print loss every n epochs
+    epochs: 1
+    # Useful debugging flag in pytorch lightning, set to True to get a single batch of training to test settings.
+    fast_dev_run: False
+    # pin images to GPU memory for fast training. This depends on GPU size and number of images.
+    preload_images: False
+
+validation:
+    # callback args
+    csv_file:
+    root_dir:
+    # Intersection over union evaluation
+    iou_threshold: 0.4
+    val_accuracy_interval: 20

everglades_dryrun_workflow.sh

@@ -3,7 +3,7 @@
 #SBATCH [email protected] 
 #SBATCH --mail-type=FAIL
 #SBATCH --gpus=a100:1
-#SBATCH --cpus-per-task=10
+#SBATCH --cpus-per-task=3
 #SBATCH --mem=200gb
 #SBATCH --time=01:30:00
 #SBATCH --partition=gpu
@@ -23,5 +23,5 @@ cd /blue/ewhite/everglades/EvergladesTools/Zooniverse
 
 snakemake --unlock
 echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] Starting Snakemake pipeline"
-snakemake --printshellcmds --keep-going --cores 10 --resources gpu=1 --rerun-incomplete --latency-wait 1 --use-conda
-echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] End"
+snakemake --printshellcmds --keep-going --cores 3 --resources gpu=1 --rerun-incomplete --latency-wait 1 --use-conda
+echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] End"

everglades_workflow.sh

@@ -23,4 +23,4 @@ cd /blue/ewhite/everglades/EvergladesTools/Zooniverse
 snakemake --unlock
 echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] Starting Snakemake pipeline"
 snakemake --printshellcmds --keep-going --cores 30 --resources gpu=1 --rerun-incomplete --latency-wait 10 --use-conda
-echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] End"
+echo "INFO [$(date "+%Y-%m-%d %H:%M:%S")] End"

predict.py

@@ -11,37 +11,6 @@
 from deepforest.utilities import boxes_to_shapefile
 
 
-def project(raster_path, boxes):
-    """
-    Convert image coordinates into a geospatial object to overlap with input image. 
-    Args:
-        raster_path: path to the raster .tif on disk. Assumed to have a valid spatial projection
-        boxes: a prediction pandas dataframe from deepforest.predict_tile()
-    Returns:
-        a geopandas dataframe with predictions in input projection.
-    """
-    with rasterio.open(raster_path) as dataset:
-        bounds = dataset.bounds
-        pixelSizeX, pixelSizeY = dataset.res
-
-    # subtract origin. Recall that numpy origin is top left! Not bottom left.
-    boxes["xmin"] = (boxes["xmin"] * pixelSizeX) + bounds.left
-    boxes["xmax"] = (boxes["xmax"] * pixelSizeX) + bounds.left
-    boxes["ymin"] = bounds.top - (boxes["ymin"] * pixelSizeY)
-    boxes["ymax"] = bounds.top - (boxes["ymax"] * pixelSizeY)
-
-    # create geometry column
-    boxes['geometry'] = boxes.apply(lambda x: shapely.geometry.box(x.xmin, x.ymin, x.xmax, x.ymax), axis=1)
-
-    # drop unnecessary columns
-    boxes = boxes.drop(columns=['xmin', 'ymin', 'xmax', 'ymax'])
-
-    # Create a GeoDataFrame
-    gdf = geopandas.GeoDataFrame(boxes, geometry='geometry')
-
-    return gdf
-
-
 def run(proj_tile_path, checkpoint_path, savedir="."):
     """Apply trained model to a drone tile"""