Adding yolo training setup

estimation/BA/BA_filtering.py

@@ -1,6 +1,7 @@
 import torch
 from BA.BA_utils import *
 
+
 def BA(iter, states, velocities, imu_meas, landmarks, landmarks_xyz, ii, time_idx, intrinsics, confidences, Sigma, V, lamda_init, poses_gt_eci, initialize=False):
 	states = states.double()
 	v = velocities.double()

estimation/BA/BA_utils.py

@@ -1,8 +1,9 @@
-import torch
 import numpy as np
+import torch
 from BA.utils import *
 from scipy.spatial import transform
-from torch_scatter import scatter_sum, scatter_mean
+from torch_scatter import scatter_mean, scatter_sum
+
 
 def proj(X, intrinsics):
     """ projection """

estimation/BA/utils.py

@@ -1,9 +1,11 @@
-import numpy as np
 import math
-import torch
+
 # from lietorch.groups import SO3, SE3
 import ipdb
+import numpy as np
+import torch
 from BA.BA_utils import *
+
 """
 Computes the Bias-Precession-Nutation matrix transforming the GCRS to the 
 CIRS intermediate reference frame. This transformation corrects for the 

estimation/errors_eval.py

@@ -1,7 +1,8 @@
-import numpy as np
 import ipdb
 import matplotlib.pyplot as plt
-
+import numpy as np
+
+
 def time_to_error():
     time_for_1 = []
     time_for_2 = []

estimation/od_pipe.py

@@ -1,15 +1,17 @@
+import gc
+import json
+import os
+
+# import pandas
+import ipdb
+import matplotlib.pyplot as plt
+import numpy as np
 import torch
+from BA.BA_filtering import BA, BA_reg
 # from lietorch.groups import SO3, SE3
 from BA.BA_utils import *
-from BA.BA_filtering import BA, BA_reg
 from BA.utils import *
-import numpy as np
-import matplotlib.pyplot as plt
-import json
-# import pandas
-import ipdb
-import os
-import gc
+
 
 def od_pipe(data, orbit_lat_long):
 

estimation/trajgen_pipe.py

@@ -1,5 +1,6 @@
-import numpy as np
 import ipdb
+import numpy as np
+
 
 class OrbitalElements:
     def __init__(self, a, e, i, Omega, omega, nu):

eval/eval_landmarks.py

@@ -1,10 +1,11 @@
+import argparse
 import os
+
 import numpy as np
-from ultralytics import YOLO
-import argparse
-from tqdm.contrib.concurrent import process_map
 import torch
-import numpy as np
+from tqdm.contrib.concurrent import process_map
+from ultralytics import YOLO
+
 
 def parse_args():
     """

labeling/label_ld.py

@@ -1,11 +1,14 @@
 import argparse
 import os
+from pathlib import Path
+
+import cv2
 import numpy as np
-import rasterio as rio
-from tqdm.contrib.concurrent import process_map
 import pyproj
+import rasterio as rio
 from PIL import Image
-import cv2
+from tqdm.contrib.concurrent import process_map
+
 
 def parse_args():
     """
@@ -33,7 +36,10 @@ def get_landmarks(landmark_path):
     Returns:
     numpy.ndarray: The loaded landmarks.
     """
-    loaded_landmarks = np.load(landmark_path)
+    if landmark_path.endswith(".csv"):
+        loaded_landmarks = np.genfromtxt(landmark_path, delimiter=",", skip_header=1)
+    else:
+        loaded_landmarks = np.load(landmark_path)
     return loaded_landmarks
 
 def get_raster_paths(raster_dir_path):
@@ -61,8 +67,8 @@ def label_raster(raster_path):
         im = src.read().transpose(1, 2, 0)
         proj = pyproj.Proj(crs)
         cxs, cys = proj(landmarks[:, 0], landmarks[:, 1])
-        lefts, bots = proj(landmarks[:, 2], landmarks[:, 3])
-        rights, tops = proj(landmarks[:, 4], landmarks[:, 5])
+        lefts, tops = proj(landmarks[:, 2], landmarks[:, 3])
+        rights, bots = proj(landmarks[:, 4], landmarks[:, 5])
         cvs, cus = src.index(cxs, cys)
         tlv, tlu = src.index(lefts, tops)
         brv, bru = src.index(rights, bots)
@@ -91,13 +97,13 @@ def label_raster(raster_path):
         cus_norm = cus_idx / im_width
         out_labels = np.stack([classes, cus_norm, cvs_norm, ws_norm, hs_norm], axis=1)
         if not os.path.exists(output_path):
-            os.makedirs(output_path)
+            Path(output_path).mkdir(parents=True, exist_ok=True)
             print('made output directory')
         if not os.path.exists(os.path.join(output_path, 'labels')):
-            os.makedirs(os.path.join(output_path, 'labels'))
+            Path(os.path.join(output_path, "labels")).mkdir(parents=True, exist_ok=True)
             print('made labels directory')
         if not os.path.exists(os.path.join(output_path, 'images')):
-            os.makedirs(os.path.join(output_path, 'images'))
+            Path(os.path.join(output_path, "images")).mkdir(parents=True, exist_ok=True)
             print('made images directory')
         file_name = os.path.basename(raster_path).split('.')[0]
         with open(os.path.join(output_path, 'labels', file_name + '.txt'), 'w') as f:

labeling/ld_downselect.py

@@ -1,6 +1,8 @@
-import numpy as np
 import argparse
 
+import numpy as np
+
+
 def parse_args():
     """
     Parse command line arguments for downselecting landmarks from Sareana.

ld_training/data_pipeline/gdrive_utils.py

@@ -0,0 +1,169 @@
+import argparse
+import logging
+import sys
+import time
+from multiprocessing import Pool
+
+from google.oauth2 import service_account
+from googleapiclient.discovery import build
+from googleapiclient.errors import HttpError
+from googleapiclient.http import MediaIoBaseDownload
+
+root = logging.getLogger()
+root.setLevel(logging.INFO)
+
+handler = logging.StreamHandler(sys.stdout)
+handler.setLevel(logging.INFO)
+formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
+handler.setFormatter(formatter)
+root.addHandler(handler)
+
+
+def parse_args():
+    """
+    Parse command line arguments for labeling Data.
+
+    Returns:
+        args (argparse.Namespace): Parsed command line arguments.
+    """
+    parser = argparse.ArgumentParser(description="Downloading Data from GDrive")
+    parser.add_argument(
+        "--credentials-path",
+        type=str,
+        required=True,
+        help="Path to credential json file",
+    )
+    parser.add_argument(
+        "--output-dir",
+        type=str,
+        required=True,
+        help="Path to save downloaded files",
+    )
+    parser.add_argument(
+        "--folders-to-download",
+        nargs='+',
+        default=[],
+        required=False,
+        help="Folder names to dowwnload, if not provided all folders will be downloaaded",
+    )
+    parsed_args = parser.parse_args()
+    return parsed_args
+
+
+def get_folder_list(drive_service):
+    try:
+        folders = []
+        page_token = None
+        while True:
+            # pylint: disable=maybe-no-member
+            response = (
+                drive_service.files()
+                .list(
+                    q="mimeType='application/vnd.google-apps.folder'",
+                    spaces="drive",
+                    fields="nextPageToken, files(id, name)",
+                    pageToken=page_token,
+                )
+                .execute()
+            )
+            logging.debug(response)
+            folders.extend(response.get("files", []))
+            page_token = response.get("nextPageToken", None)
+            if page_token is None:
+                break
+    except HttpError as error:
+        logging.error(f"An error occurred: {error}")
+        folders = None
+    return folders
+
+
+def get_tiff_file_list(drive_service, folder_id):
+    """Search file in drive location
+
+    Load pre-authorized user credentials from the environment.
+    TODO(developer) - See https://developers.google.com/identity
+    for guides on implementing OAuth2 for the application.
+    """
+
+    query_string = "mimeType='image/tiff'"
+    query_string += " and '" + folder_id + "' in parents"
+
+    logging.debug(query_string)
+
+    try:
+        files = []
+        page_token = None
+        while True:
+            # pylint: disable=maybe-no-member
+            response = (
+                drive_service.files()
+                .list(
+                    q=query_string,
+                    spaces="drive",
+                    fields="nextPageToken, files(id, name)",
+                    pageToken=page_token,
+                )
+                .execute()
+            )
+            logging.debug(response)
+            files.extend(response.get("files", []))
+            page_token = response.get("nextPageToken", None)
+            if page_token is None:
+                break
+    except HttpError as error:
+        logging.error(f"An error occurred: {error}")
+        files = None
+    return files
+
+
+def download_file_from_google_drive(drive_service, file_id, destination):
+    start = time.perf_counter()
+    request = drive_service.files().get_media(fileId=file_id)
+    with open(destination, "wb") as f:
+        downloader = MediaIoBaseDownload(f, request)
+        done = False
+        while done is False:
+            status, done = downloader.next_chunk()
+            logging.debug("Download %d%%." % int(status.progress() * 100))
+    download_time_seconds = time.perf_counter() - start
+    logging.info(
+        f"file id: {file_id} destination: {destination} download_time_seconds {download_time_seconds}"
+    )
+
+
+def multi_process_download(files):
+    result = []
+    pool = Pool(processes=200)
+    res = pool.starmap(download_file_from_google_drive, files, chunksize=10)
+    result.extend(res)
+    pool.close()
+    pool.join()
+    logging.info(result)
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    logging.info(f"Args provided: {args}")
+    creds = service_account.Credentials.from_service_account_file(
+        args.credentials_path,
+        scopes=["https://www.googleapis.com/auth/drive"],
+    )
+    drive_service = build("drive", "v3", credentials=creds)
+    folders = get_folder_list(drive_service)
+    logging.info(folders)
+    filtered_folders = folders
+    if len(args.folders_to_download) > 0:
+        filtered_folders = [f for f in folders if f["name"] in args.folders_to_download]
+    for folder in filtered_folders:
+        files = get_tiff_file_list(drive_service, folder["id"])
+        destination_dir = args.output_dir
+        if files:
+            mp_list = []
+            for file in files:
+                mp_list.append(
+                    (drive_service, file["id"], f"{destination_dir}/{file['name']}")
+                )
+            logging.info(
+                f"Number of files in folder {folder['name']}: {len(mp_list)}"
+            )
+            multi_process_download(mp_list)

ld_training/yolo/datasets/12R.yaml

@@ -0,0 +1,5 @@
+# dataset root dir
+path: /home/giriman/temp/
+train: label_12R/images
+val: label_12R_val/images
+nc: 500

ld_training/yolo/datasets/eedl_cmds.txt

@@ -0,0 +1,12 @@
+train: 
+- python eedl.py -g 12R -i 2020 -f 2023 -s 150 -m 2000 -o sentinel_images_12R -se s2 --gdrive True
+- python eedl.py -g 12R -i 2015 -f 2017 -s 150 -m 2000 -o sentinel_images_12R -se s2 --gdrive True
+
+val: 
+- python eedl.py -g 12R -i 2018 -f 2019 -s 150 -m 200 -o sentinel_images_12R_val -se s2 --gdrive True
+
+train: python3 label_ld.py --landmark_path /home/giriman/CMU/rexlab/sat/VINSat/sim/landmark_csvs/12R_top_salient.csv --raster_dir_path /home/giriman/CMU/rexlab/sat/data/sentinel/12R/ --output_dir_path /home/giriman/CMU/rexlab/sat/data/sentinel/label_12R/ -v -c
+
+val: python3 label_ld.py --landmark_path /home/giriman/CMU/rexlab/sat/VINSat/sim/landmark_csvs/12R_top_salient.csv --raster_dir_path /home/giriman/CMU/rexlab/sat/data/sentinel/12R_val/ --output_dir_path /home/giriman/CMU/rexlab/sat/data/sentinel/label_12R_val/ -v -c
+
+python3 gdrive_utils.py --credentials-path /home/giriman/gcp/credentials/ee-cmu-vinsat-giria-46a18fe24bfb.json --output-dir /home/giriman/temp --folders-to-download sentinel_images_12R_val

ld_training/yolo/utils/plot_results.py

@@ -0,0 +1,6 @@
+from ultralytics.utils.plotting import plot_results
+
+if __name__ == "__main__":
+    plot_results(
+        "/home/giriman/CMU/rexlab/sat/VINSat/ld_training/yolo/runs/detect/17R_n1342_top_salient_mse/results.csv")
+