update installation

.gitattributes

@@ -1,2 +1,2 @@
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.weights filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.weights filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -1,22 +1,22 @@
-.pth
-*.mp4
-*.jpg
-*.png
-*.swp
-**/utils/gs/
-**/.DS_Store
-
-**/.idea
-**/*.egg-info/
-**/__pycache__
-**/build
-*.py[cod]
-**/bin/*
-*.log
-*.error
-*.a
-*.so
-*.so.2
-.ipynb_checkpoints
-.python-version
-*/gs/*
+.pth
+*.mp4
+*.jpg
+*.png
+*.swp
+**/utils/gs/
+**/.DS_Store
+
+**/.idea
+**/*.egg-info/
+**/__pycache__
+**/build
+*.py[cod]
+**/bin/*
+*.log
+*.error
+*.a
+*.so
+*.so.2
+.ipynb_checkpoints
+.python-version
+*/gs/*

CVC-YOLOv3/README.md

@@ -1,72 +1,72 @@
-### Description
-
-The repo is forked from https://github.com/ultralytics/yolov3 and contains inference and training code for YOLOv3 in PyTorch.
-
-### Setup
-Make sure to have followed `setup-gcloud.md` and be logged into the corresponding GCloud instance. In particular, make sure to have run the `setup.py` command as directed there.
-
-Then run the following, note that the generate data script has to have links that are the gs: version and not the 'storage.api' version:
-```
-source /home/mit-dut-driverless-internal/venvs/cv/bin/activate
-# log in through the browser on your local machine
-gcloud auth application-default login 
-# Note that the csv_uri must have at least 100 images in it and it must have the gs: locations referenced rather than their local locations
-python generate_dataset_csvs.py --csv_uris \
-	gs://mit-dut-driverless-internal/data-labels/fullFrameTest.csv \
-	--output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/fullFrameNonSquareTest/
-
-python train.py --model_cfg model_cfgs/yolov3_80class.cfg \
-	--validate_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-test.csv \
-	--train_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-train.csv \
-	--output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/quickstart/ \
-	--study_name color_80class \
-	--weights_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/92da791a96de485895ea219f7035c2aa/36.weights 2>&1 | tee results/color_416_baseline.log
-```
-
-The same command in one line:
-```
-python train.py --model_cfg model_cfgs/yolov3_80class.cfg --validate_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-test.csv --train_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-train.csv --output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline --study_name color_416_baseline 2>&1 | tee logs/color_416_baseline.log
-```
-
-Back-up the logs:
-```
-gsutil cp logs/color_416_baseline.log gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline/color_416_baseline.log
-```
-
-To run detection on one image, and visualize the results:
-```
-python detect.py --weights_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/sample-yolov3.weights --image_uri gs://mit-dut-driverless-external/HiveAIRound2/vid_38_frame_956.jpg --output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/fullFrameNonSquareTest2/vid38f956.jpg --img_width 2048 --img_height 1536 --model_cfg model_cfgs/yolov3_80class_fullFrame.cfg
-```
-To run detection on multiple images of your choosing, write a bash file similar to:
-```
-run_scripts/run_detect.sh
-```
-And to copy the visualizations to a local machine (with GCloud Client installed): `gsutil -m cp -r gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline/visualization .`
-
-You can also add the following to your ~/.bashrc to make things easier:
-```
-cd /home/mit-dut-driverless-internal/cv-cone-town/vectorized-yolov3/
-source ../../venvs/cv/bin/activate
-```
-
-You can create a video from a list of extracted frames from a video using:
-```
-run_scripts/make_video.sh
-```
-You'll need to adjust the hardcoded paths at the top of the file.
-
-To-Do:
-Splits:
-    - Get splits working with online processing, re-train models
-    - Understand why splits work so well
-    - Add splits to Xavier processing module
-- Make experiment folders readable
-
-### Bookmarked Models
-Color Full-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/92da791a96de485895ea219f7035c2aa/36.weights
-Black and White Full-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/2c5485a8ee6847808459f54cac50ae8e/64.weights
-
-Color Split-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/56070f8a1e454b2383c12d0fec37e3dc/104.weights
-
-Black and White Split-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/701e0c805b4d4052a1798a4d9c3c5914/68.weights
-
+### Description
+
+The repo is forked from https://github.com/ultralytics/yolov3 and contains inference and training code for YOLOv3 in PyTorch.
+
+### Setup
+Make sure to have followed `setup-gcloud.md` and be logged into the corresponding GCloud instance. In particular, make sure to have run the `setup.py` command as directed there.
+
+Then run the following, note that the generate data script has to have links that are the gs: version and not the 'storage.api' version:
+```
+source /home/mit-dut-driverless-internal/venvs/cv/bin/activate
+# log in through the browser on your local machine
+gcloud auth application-default login 
+# Note that the csv_uri must have at least 100 images in it and it must have the gs: locations referenced rather than their local locations
+python generate_dataset_csvs.py --csv_uris \
+	gs://mit-dut-driverless-internal/data-labels/fullFrameTest.csv \
+	--output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/fullFrameNonSquareTest/
+
+python train.py --model_cfg model_cfgs/yolov3_80class.cfg \
+	--validate_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-test.csv \
+	--train_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-train.csv \
+	--output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/quickstart/ \
+	--study_name color_80class \
+	--weights_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/92da791a96de485895ea219f7035c2aa/36.weights 2>&1 | tee results/color_416_baseline.log
+```
+
+The same command in one line:
+```
+python train.py --model_cfg model_cfgs/yolov3_80class.cfg --validate_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-test.csv --train_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/hive-0-1-2-train.csv --output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline --study_name color_416_baseline 2>&1 | tee logs/color_416_baseline.log
+```
+
+Back-up the logs:
+```
+gsutil cp logs/color_416_baseline.log gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline/color_416_baseline.log
+```
+
+To run detection on one image, and visualize the results:
+```
+python detect.py --weights_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/sample-yolov3.weights --image_uri gs://mit-dut-driverless-external/HiveAIRound2/vid_38_frame_956.jpg --output_uri gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/fullFrameNonSquareTest2/vid38f956.jpg --img_width 2048 --img_height 1536 --model_cfg model_cfgs/yolov3_80class_fullFrame.cfg
+```
+To run detection on multiple images of your choosing, write a bash file similar to:
+```
+run_scripts/run_detect.sh
+```
+And to copy the visualizations to a local machine (with GCloud Client installed): `gsutil -m cp -r gs://mit-dut-driverless-internal/vectorized-yolov3-training/january-experiments/color_416_baseline/visualization .`
+
+You can also add the following to your ~/.bashrc to make things easier:
+```
+cd /home/mit-dut-driverless-internal/cv-cone-town/vectorized-yolov3/
+source ../../venvs/cv/bin/activate
+```
+
+You can create a video from a list of extracted frames from a video using:
+```
+run_scripts/make_video.sh
+```
+You'll need to adjust the hardcoded paths at the top of the file.
+
+To-Do:
+Splits:
+    - Get splits working with online processing, re-train models
+    - Understand why splits work so well
+    - Add splits to Xavier processing module
+- Make experiment folders readable
+
+### Bookmarked Models
+Color Full-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/92da791a96de485895ea219f7035c2aa/36.weights
+Black and White Full-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/2c5485a8ee6847808459f54cac50ae8e/64.weights
+
+Color Split-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/56070f8a1e454b2383c12d0fec37e3dc/104.weights
+
+Black and White Split-Frame: gs://mit-dut-driverless-internal/vectorized-yolov3-training/december-experiments/701e0c805b4d4052a1798a4d9c3c5914/68.weights
+

CVC-YOLOv3/csrc/ROIAlign.h

@@ -1,46 +1,46 @@
-// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-#pragma once
-
-#include "cpu/vision.h"
-
-#ifdef WITH_CUDA
-#include "cuda/vision.h"
-#endif
-
-// Interface for Python
-at::Tensor ROIAlign_forward(const at::Tensor& input,
-                            const at::Tensor& rois,
-                            const float spatial_scale,
-                            const int pooled_height,
-                            const int pooled_width,
-                            const int sampling_ratio) {
-  if (input.type().is_cuda()) {
-#ifdef WITH_CUDA
-    return ROIAlign_forward_cuda(input, rois, spatial_scale, pooled_height, pooled_width, sampling_ratio);
-#else
-    AT_ERROR("Not compiled with GPU support");
-#endif
-  }
-  return ROIAlign_forward_cpu(input, rois, spatial_scale, pooled_height, pooled_width, sampling_ratio);
-}
-
-at::Tensor ROIAlign_backward(const at::Tensor& grad,
-                             const at::Tensor& rois,
-                             const float spatial_scale,
-                             const int pooled_height,
-                             const int pooled_width,
-                             const int batch_size,
-                             const int channels,
-                             const int height,
-                             const int width,
-                             const int sampling_ratio) {
-  if (grad.type().is_cuda()) {
-#ifdef WITH_CUDA
-    return ROIAlign_backward_cuda(grad, rois, spatial_scale, pooled_height, pooled_width, batch_size, channels, height, width, sampling_ratio);
-#else
-    AT_ERROR("Not compiled with GPU support");
-#endif
-  }
-  AT_ERROR("Not implemented on the CPU");
-}
-
+// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+#pragma once
+
+#include "cpu/vision.h"
+
+#ifdef WITH_CUDA
+#include "cuda/vision.h"
+#endif
+
+// Interface for Python
+at::Tensor ROIAlign_forward(const at::Tensor& input,
+                            const at::Tensor& rois,
+                            const float spatial_scale,
+                            const int pooled_height,
+                            const int pooled_width,
+                            const int sampling_ratio) {
+  if (input.type().is_cuda()) {
+#ifdef WITH_CUDA
+    return ROIAlign_forward_cuda(input, rois, spatial_scale, pooled_height, pooled_width, sampling_ratio);
+#else
+    AT_ERROR("Not compiled with GPU support");
+#endif
+  }
+  return ROIAlign_forward_cpu(input, rois, spatial_scale, pooled_height, pooled_width, sampling_ratio);
+}
+
+at::Tensor ROIAlign_backward(const at::Tensor& grad,
+                             const at::Tensor& rois,
+                             const float spatial_scale,
+                             const int pooled_height,
+                             const int pooled_width,
+                             const int batch_size,
+                             const int channels,
+                             const int height,
+                             const int width,
+                             const int sampling_ratio) {
+  if (grad.type().is_cuda()) {
+#ifdef WITH_CUDA
+    return ROIAlign_backward_cuda(grad, rois, spatial_scale, pooled_height, pooled_width, batch_size, channels, height, width, sampling_ratio);
+#else
+    AT_ERROR("Not compiled with GPU support");
+#endif
+  }
+  AT_ERROR("Not implemented on the CPU");
+}
+

CVC-YOLOv3/csrc/ROIPool.h

@@ -1,48 +1,48 @@
-// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
-#pragma once
-
-#include "cpu/vision.h"
-
-#ifdef WITH_CUDA
-#include "cuda/vision.h"
-#endif
-
-
-std::tuple<at::Tensor, at::Tensor> ROIPool_forward(const at::Tensor& input,
-                                const at::Tensor& rois,
-                                const float spatial_scale,
-                                const int pooled_height,
-                                const int pooled_width) {
-  if (input.type().is_cuda()) {
-#ifdef WITH_CUDA
-    return ROIPool_forward_cuda(input, rois, spatial_scale, pooled_height, pooled_width);
-#else
-    AT_ERROR("Not compiled with GPU support");
-#endif
-  }
-  AT_ERROR("Not implemented on the CPU");
-}
-
-at::Tensor ROIPool_backward(const at::Tensor& grad,
-                                 const at::Tensor& input,
-                                 const at::Tensor& rois,
-                                 const at::Tensor& argmax,
-                                 const float spatial_scale,
-                                 const int pooled_height,
-                                 const int pooled_width,
-                                 const int batch_size,
-                                 const int channels,
-                                 const int height,
-                                 const int width) {
-  if (grad.type().is_cuda()) {
-#ifdef WITH_CUDA
-    return ROIPool_backward_cuda(grad, input, rois, argmax, spatial_scale, pooled_height, pooled_width, batch_size, channels, height, width);
-#else
-    AT_ERROR("Not compiled with GPU support");
-#endif
-  }
-  AT_ERROR("Not implemented on the CPU");
-}
-
-
-
+// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
+#pragma once
+
+#include "cpu/vision.h"
+
+#ifdef WITH_CUDA
+#include "cuda/vision.h"
+#endif
+
+
+std::tuple<at::Tensor, at::Tensor> ROIPool_forward(const at::Tensor& input,
+                                const at::Tensor& rois,
+                                const float spatial_scale,
+                                const int pooled_height,
+                                const int pooled_width) {
+  if (input.type().is_cuda()) {
+#ifdef WITH_CUDA
+    return ROIPool_forward_cuda(input, rois, spatial_scale, pooled_height, pooled_width);
+#else
+    AT_ERROR("Not compiled with GPU support");
+#endif
+  }
+  AT_ERROR("Not implemented on the CPU");
+}
+
+at::Tensor ROIPool_backward(const at::Tensor& grad,
+                                 const at::Tensor& input,
+                                 const at::Tensor& rois,
+                                 const at::Tensor& argmax,
+                                 const float spatial_scale,
+                                 const int pooled_height,
+                                 const int pooled_width,
+                                 const int batch_size,
+                                 const int channels,
+                                 const int height,
+                                 const int width) {
+  if (grad.type().is_cuda()) {
+#ifdef WITH_CUDA
+    return ROIPool_backward_cuda(grad, input, rois, argmax, spatial_scale, pooled_height, pooled_width, batch_size, channels, height, width);
+#else
+    AT_ERROR("Not compiled with GPU support");
+#endif
+  }
+  AT_ERROR("Not implemented on the CPU");
+}
+
+
+