diff --git a/.gitignore b/.gitignore
index 37b3e6d..bec4f9f 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,2 +1,5 @@
 .DS_Store
-/Data/
+/Data/fsoco_segmentation_train
+/Segmentation/Yolact_minimal
+/SampleData/driverless.mp4
+*.txt
diff --git a/Data/README.md b/Data/README.md
new file mode 100644
index 0000000..fb383ce
--- /dev/null
+++ b/Data/README.md
@@ -0,0 +1 @@
+Clone FS coco data into here
diff --git a/Projects/00Template.md b/Projects/00Template.md
new file mode 100644
index 0000000..f2e6795
--- /dev/null
+++ b/Projects/00Template.md
@@ -0,0 +1,13 @@
+# Project Title
+
+## Active Members
+
+## Design Review Dates/Timeline
+
+## Introduction
+
+## Overview of the Problem
+
+## Steps in the project
+
+## Suggested direction
\ No newline at end of file
diff --git a/Projects/ControlPlanning.md b/Projects/ControlPlanning.md
new file mode 100644
index 0000000..5586123
--- /dev/null
+++ b/Projects/ControlPlanning.md
@@ -0,0 +1,30 @@
+# Control Planning
+
+## Active Members
+1. Ray Zou
+
+## Design Review Dates/Timeline
+1. PDR November 3rd
+
+## Introduction
+Control input planning is a vehicle dynamics problem to generate the
+optimal set of vehicle dynamics inputs needed to follow a
+predetermined path. Other projects, such as path planning will already
+generate a path that must be followed. 
+
+## Overview of the Problem
+Using and working with the VD and Simulations team, there will be a
+program that exists that can use 3 degrees of freedom (brake,
+throttle, steering) to estimate the change of position over the next
+given amount of time. Using this simulation and a given path, you must
+identify what inputs are required to follow the given path. 
+
+## Steps in the project
+1. Understand the file format for the given path
+1. Understand the capabilities of the vehicle dynamics simulations
+1. Literature review over methods to follow a line
+1. Design of a control loop
+1. Implementation of input selecting for a short period of time
+1. Optimization
+
+## Suggested direction
diff --git a/Projects/IMUDataProcessing.md b/Projects/IMUDataProcessing.md
new file mode 100644
index 0000000..b3d0acc
--- /dev/null
+++ b/Projects/IMUDataProcessing.md
@@ -0,0 +1,30 @@
+# IMU Data Processing
+
+## Active Members
+1. Saahith Veeramaneni
+
+## Design Review Dates/Timeline
+1. Literature review by November 3rd
+1. PDR by November 10th
+1. Theoretical codebase, Thanksgiving
+
+## Introduction
+Due to the low compute on the car, it is not possible to use multiple
+cameras so this project will explore using structure from motion (sfm)
+to create binocular views.
+
+## Overview of the Problem
+Having multiple perspectives on a car allows the creation of a
+disparity map from an image which can be used to estimate depth if the
+exact distance between the cameras is well understood. Thus, we must
+explore if we can achieve high enough accuracy from the IMU to use sfm
+in the data processing pipeline.
+
+## Steps in the project
+1. Literature review
+1. Understanding compute requirements
+1. Draft of the code if it is feasible
+
+## Suggested direction
+1. Begin with a long literature review and exploring its use on other
+   Formula Student teams and in industry
diff --git a/Projects/Localization.md b/Projects/Localization.md
new file mode 100644
index 0000000..14d79f7
--- /dev/null
+++ b/Projects/Localization.md
@@ -0,0 +1,29 @@
+# Localization
+
+## Active Members
+1. Parker Costa
+1. Hansika Nerusa
+1. Chanchal Mukeshsingh
+
+## Design Review Dates/Timeline
+1. PDR by november 4th
+1. Creation of an independent working system by end of first quarter
+
+## Introduction
+Localization deals with the creation of both a global and local map
+based on camera inputs so that the car can understand how its moving
+and its surroundings. 
+
+## Overview of the Problem
+By being fed in raw data from the IMU, depth estimation, and cone
+segmentation, the project must identify where cones are and create a
+global map overtime by processing raw data and estimating where cones
+are even if the data is noisy.
+
+## Steps in the project
+1. Literature review and PDR
+1. Communication and finalization of input and output data formats 
+1. Code base developed in a virtual test bench
+1. Code review
+
+## Suggested direction
diff --git a/Projects/PathPlanning.md b/Projects/PathPlanning.md
new file mode 100644
index 0000000..c117d41
--- /dev/null
+++ b/Projects/PathPlanning.md
@@ -0,0 +1,28 @@
+# Path Planning
+
+## Active Members
+1. Suhani Agarwal
+1. Nathan Yee
+1. Aedan Benavides
+
+## Design Review Dates/Timeline
+1. PDR by November 3rd
+1. Finished by end of winter quarter
+
+## Introduction
+The project focuses on identifying a path from the map data by
+using the cones to find the optimal path to move through the cones.
+
+## Overview of the Problem
+When raw map data is presented it'll be done in the form of a
+coordinate system on a 2D plane. To generate a path to plan a route
+through, a polygon and direction representation of the path must be identified.
+
+## Steps in the project
+1. Literature review
+1. PDR
+1. Triangulation methods
+1. Data format standardization
+
+## Suggested direction
+- Unknown
diff --git a/Projects/Yolact.md b/Projects/Yolact.md
new file mode 100644
index 0000000..b074a3a
--- /dev/null
+++ b/Projects/Yolact.md
@@ -0,0 +1,33 @@
+# YOLACT
+
+## Active Members
+1. Abishek Adari
+1. Dylan Price
+1.Jaisree D. RaviKumar
+
+## Design Review Dates/Timeline
+1. Read the paper and PDR  November 3rd
+1. Implementation and training of YOLACT edge by November 24th
+1. Benchmarking and optimization for edge devices, end of quarter
+
+## Introduction
+Identifying cones for processing is a difficult problem and YOLACT
+performs semantic segmentation to identify and classify each
+individual pixel among their relevant classes.
+
+## Overview of the Problem
+The self driving car must know the bounds it can travel, and the
+creation of the bounds must involve identifying where in an image the
+cone is. We do this using an implementation of the YOLACTEdge model
+which will classify every pixel among all possibilities. We target at
+least 30FPS when running alone on a Jetson Orin Nano.
+
+## Steps in the project
+1. Read the papers
+1. Base reimplementation of the project
+1. Data processing
+1. Training
+1. Evaluation
+
+## Suggested direction
+Read the YOLACTEdge, Yolact, and YOLACT++ papers.
diff --git a/SampleData/README.md b/SampleData/README.md
new file mode 100644
index 0000000..3b958dd
--- /dev/null
+++ b/SampleData/README.md
@@ -0,0 +1,2 @@
+Download the youtube video here:
+https://www.youtube.com/watch?v=o5vES5QaeiQ
diff --git a/Segmentation/README.md b/Segmentation/README.md
new file mode 100644
index 0000000..340244e
--- /dev/null
+++ b/Segmentation/README.md
@@ -0,0 +1,4 @@
+Cloen this repo into this directory to run segmentation:
+https://github.com/feiyuhuahuo/Yolact_minimal
+
+Move the FS 
diff --git a/Segmentation/__pycache__/convertSupervisely2Coco.cpython-312.pyc b/Segmentation/__pycache__/convertSupervisely2Coco.cpython-312.pyc
new file mode 100644
index 0000000..7a18da5
Binary files /dev/null and b/Segmentation/__pycache__/convertSupervisely2Coco.cpython-312.pyc differ
diff --git a/Segmentation/convertSupervisely2Coco.py b/Segmentation/convertSupervisely2Coco.py
new file mode 100644
index 0000000..bcf98a0
--- /dev/null
+++ b/Segmentation/convertSupervisely2Coco.py
@@ -0,0 +1,143 @@
+import json
+import os
+import glob
+import base64
+import zlib
+import numpy as np
+from pathlib import Path
+import cv2
+from io import BytesIO
+
+def decodeSuperviselyBitmap(bitmapData, origin, imgHeight, imgWidth):
+    data = base64.b64decode(bitmapData['data'])
+    pngData = zlib.decompress(data)
+
+    maskImg = cv2.imdecode(np.frombuffer(pngData, dtype=np.uint8), cv2.IMREAD_GRAYSCALE)
+
+    if maskImg is None:
+        return None
+
+    mask = (maskImg > 0).astype(np.uint8)
+
+    fullMask = np.zeros((imgHeight, imgWidth), dtype=np.uint8)
+    y1, x1 = origin[1], origin[0]
+    y2, x2 = min(y1 + mask.shape[0], imgHeight), min(x1 + mask.shape[1], imgWidth)
+    fullMask[y1:y2, x1:x2] = mask[:y2-y1, :x2-x1]
+    return fullMask
+
+def maskToRle(mask):
+    pixels = mask.flatten()
+    pixels = np.concatenate([[0], pixels, [0]])
+    runs = np.where(pixels[1:] != pixels[:-1])[0] + 1
+    runs[1::2] -= runs[::2]
+    return runs.tolist()
+
+def maskToBbox(mask):
+    rows = np.any(mask, axis=1)
+    cols = np.any(mask, axis=0)
+    if not rows.any() or not cols.any():
+        return [0, 0, 0, 0]
+    ymin, ymax = np.where(rows)[0][[0, -1]]
+    xmin, xmax = np.where(cols)[0][[0, -1]]
+    return [int(xmin), int(ymin), int(xmax - xmin + 1), int(ymax - ymin + 1)]
+
+def convertSupervisely2Coco(dataRoot, outputPath):
+    metaPath = os.path.join(dataRoot, 'meta.json')
+    with open(metaPath, 'r') as f:
+        meta = json.load(f)
+
+    segClasses = [c for c in meta['classes'] if c['shape'] == 'bitmap']
+    classIdToCocoId = {c['id']: idx + 1 for idx, c in enumerate(segClasses)}
+
+    cocoData = {
+        'images': [],
+        'annotations': [],
+        'categories': [{'id': idx + 1, 'name': c['title'], 'supercategory': 'cone'}
+                       for idx, c in enumerate(segClasses)]
+    }
+
+    print(f"Found {len(segClasses)} segmentation classes:")
+    for cat in cocoData['categories']:
+        print(f"  {cat['id']}: {cat['name']}")
+
+    teamDirs = [d for d in os.listdir(dataRoot) if os.path.isdir(os.path.join(dataRoot, d))]
+
+    imageId = 0
+    annotationId = 0
+
+    for teamDir in sorted(teamDirs):
+        annDir = os.path.join(dataRoot, teamDir, 'ann')
+        imgDir = os.path.join(dataRoot, teamDir, 'img')
+
+        if not os.path.exists(annDir) or not os.path.exists(imgDir):
+            continue
+
+        annFiles = glob.glob(os.path.join(annDir, '*.json'))
+        print(f"\nProcessing {teamDir}: {len(annFiles)} images", flush=True)
+
+        for idx, annFile in enumerate(sorted(annFiles)):
+            if idx % 20 == 0:
+                print(f"  {teamDir}: {idx}/{len(annFiles)}", flush=True)
+            with open(annFile, 'r') as f:
+                ann = json.load(f)
+
+            imgFilename = os.path.basename(annFile).replace('.json', '')
+            imgPath = os.path.join(teamDir, 'img', imgFilename)
+
+            cocoData['images'].append({
+                'id': imageId,
+                'file_name': imgPath,
+                'height': ann['size']['height'],
+                'width': ann['size']['width']
+            })
+
+            for obj in ann['objects']:
+                if obj['geometryType'] != 'bitmap':
+                    continue
+
+                classId = obj['classId']
+                if classId not in classIdToCocoId:
+                    continue
+
+                cocoId = classIdToCocoId[classId]
+
+                fullMask = decodeSuperviselyBitmap(obj['bitmap'], obj['bitmap']['origin'],
+                                                  ann['size']['height'], ann['size']['width'])
+
+                if fullMask is None:
+                    continue
+
+                bbox = maskToBbox(fullMask)
+                area = int(np.sum(fullMask))
+
+                if area == 0:
+                    continue
+
+                rle = maskToRle(fullMask)
+
+                cocoData['annotations'].append({
+                    'id': annotationId,
+                    'image_id': imageId,
+                    'category_id': cocoId,
+                    'segmentation': [rle],
+                    'area': area,
+                    'bbox': bbox,
+                    'iscrowd': 0
+                })
+                annotationId += 1
+
+            imageId += 1
+
+    print(f"\n{'='*60}")
+    print(f"Total images: {len(cocoData['images'])}")
+    print(f"Total annotations: {len(cocoData['annotations'])}")
+
+    with open(outputPath, 'w') as f:
+        json.dump(cocoData, f)
+
+    print(f"Saved to: {outputPath}")
+
+if __name__ == '__main__':
+    dataRoot = 'Data/fsoco_segmentation_train'
+    outputPath = 'Data/fsoco_segmentation_train/train_coco.json'
+    convertSupervisely2Coco(dataRoot, outputPath)
diff --git a/Segmentation/detectCones.py b/Segmentation/detectCones.py
new file mode 100755
index 0000000..2386f63
--- /dev/null
+++ b/Segmentation/detectCones.py
@@ -0,0 +1,39 @@
+#!/usr/bin/env python3
+import sys
+import os
+import subprocess
+
+def main():
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+    yolact_dir = os.path.join(script_dir, 'yolact_edge')
+
+    args = sys.argv[1:]
+
+    default_args = [
+        '--config=yolact_edge_mobilenetv2_cone_config',
+        '--score_threshold=0.3',
+        '--top_k=100',
+    ]
+
+    provided_flags = {arg.split('=')[0].lstrip('-') for arg in args if '=' in arg or arg.startswith('--')}
+
+    final_args = []
+    for default_arg in default_args:
+        flag = default_arg.split('=')[0].lstrip('-')
+        if flag not in provided_flags and not any(arg.startswith('--' + flag) for arg in args):
+            final_args.append(default_arg)
+
+    final_args.extend(args)
+
+    cmd = [sys.executable, 'eval.py'] + final_args
+
+    print(f"Running inference in {yolact_dir}")
+    print(f"Command: {' '.join(cmd)}\n")
+
+    os.chdir(yolact_dir)
+
+    result = subprocess.run(cmd)
+    sys.exit(result.returncode)
+
+if __name__ == '__main__':
+    main()
diff --git a/Segmentation/trainCones.py b/Segmentation/trainCones.py
new file mode 100755
index 0000000..dbbe757
--- /dev/null
+++ b/Segmentation/trainCones.py
@@ -0,0 +1,43 @@
+#!/usr/bin/env python3
+import sys
+import os
+import subprocess
+
+def main():
+    script_dir = os.path.dirname(os.path.abspath(__file__))
+    yolact_dir = os.path.join(script_dir, 'yolact_edge')
+
+    args = sys.argv[1:]
+
+    default_args = [
+        '--config=yolact_edge_mobilenetv2_cone_config',
+        '--batch_size=4',
+        '--save_folder=weights/',
+        '--validation_epoch=2',
+        '--save_interval=5000',
+        '--max_checkpoints=10',
+        '--num_workers=0',
+    ]
+
+    provided_flags = {arg.split('=')[0].lstrip('-') for arg in args if '=' in arg or arg.startswith('--')}
+
+    final_args = []
+    for default_arg in default_args:
+        flag = default_arg.split('=')[0].lstrip('-')
+        if flag not in provided_flags and not any(arg.startswith('--' + flag) for arg in args):
+            final_args.append(default_arg)
+
+    final_args.extend(args)
+
+    cmd = [sys.executable, 'train.py'] + final_args
+
+    print(f"Running training in {yolact_dir}")
+    print(f"Command: {' '.join(cmd)}\n")
+
+    os.chdir(yolact_dir)
+
+    result = subprocess.run(cmd)
+    sys.exit(result.returncode)
+
+if __name__ == '__main__':
+    main()
diff --git a/viewData.py b/viewData.py
new file mode 100644
index 0000000..87953be
--- /dev/null
+++ b/viewData.py
@@ -0,0 +1,29 @@
+import json
+import base64
+import io
+import zlib
+import numpy as np
+from PIL import Image
+import matplotlib.pyplot as plt
+
+with open('Data/fsoco_segmentation_train/epflrt/ann/amz_00825.png.json') as f:
+    ann = json.load(f)
+
+img = Image.open('Data/fsoco_segmentation_train/epflrt/img/amz_00825.png')
+fullMask = np.zeros((ann['size']['height'], ann['size']['width']), dtype=np.uint8)
+
+for obj in ann['objects']:
+    if obj['geometryType'] == 'bitmap':
+        bitmapData = base64.b64decode(obj['bitmap']['data'])
+
+        decompressed = zlib.decompress(bitmapData)
+        mask = Image.open(io.BytesIO(decompressed)).convert('L')
+
+        origin = obj['bitmap']['origin']
+        maskArray = np.array(mask)
+        fullMask[origin[1]:origin[1]+mask.height, origin[0]:origin[0]+mask.width] = np.maximum(fullMask[origin[1]:origin[1]+mask.height, origin[0]:origin[0]+mask.width], maskArray)
+
+plt.figure(figsize=(15, 8))
+plt.imshow(img)
+plt.imshow(fullMask, alpha=0.5)
+plt.show()