OpenSplat/input_data.cpp at OpenSplatLibrary · NewChromantics/OpenSplat · GitHub

History

474 lines (384 loc) · 13.8 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

#include <filesystem>

#include <nlohmann/json.hpp>

#include "input_data.hpp"

#include "cv_utils.hpp"

#include "trainer.hpp" // NoCameraException

#include "nerfstudio.hpp"

namespace fs = std::filesystem;

using namespace torch::indexing;

using json = nlohmann::json;

namespace cm{ InputData inputDataFromColmap(const std::string &projectRoot, const std::string& imageSourcePath); }

namespace osfm { InputData inputDataFromOpenSfM(const std::string &projectRoot); }

namespace omvg { InputData inputDataFromOpenMVG(const std::string &projectRoot); }

InputData inputDataFromX(const std::string &projectRoot, const std::string& colmapImageSourcePath,bool CenterAndNormalisePoints,bool AddCameras)

{

fs::path root(projectRoot);

if (fs::exists(root / "transforms.json")){

return ns::inputDataFromNerfStudio(projectRoot,CenterAndNormalisePoints,AddCameras);

}else if (fs::exists(root / "sparse") || fs::exists(root / "cameras.bin")){

return cm::inputDataFromColmap(projectRoot, colmapImageSourcePath);

}else if (fs::exists(root / "reconstruction.json")){

return osfm::inputDataFromOpenSfM(projectRoot);

}else if (fs::exists(root / "opensfm" / "reconstruction.json")){

return osfm::inputDataFromOpenSfM((root / "opensfm").string());

}else if (fs::exists(root / "sfm_data.json")){

return omvg::inputDataFromOpenMVG((root).string());

}

else{

throw std::runtime_error("Invalid project folder (must be either a colmap or nerfstudio or openmvg project folder)");

}

Camera::Camera(CameraIntrinsics intrinsics,const torch::Tensor &camToWorld,std::filesystem::path cameraImageFilename) :

intrinsics(intrinsics),

camToWorld(camToWorld),

cameraImagePath(cameraImageFilename)

{

}

void CameraIntrinsics::RemoveDistortionParameters()

{

k1 = 0;

k2 = 0;

k3 = 0;

p1 = 0;

p2 = 0;

}

void CameraIntrinsics::ScaleTo(int Width,int Height)

{

// allowing x & y scale is going to mess up distortion parameters

// this would be fine for forward rendering (which doesnt distort)

// if there are distortion values here for record of the original camera

// they should really have been zeroed out already if they no longer apply

// to the image they're attached to

if ( HasDistortionParameters() )

throw std::runtime_error("Cannot scale CameraIntrinsics which have distortion parameters");

if ( Width <= 0 || Height <= 0 )

{

std::stringstream Error;

Error << "Cannot scale CameraIntrinsics to " << Width << "x" << Height;

throw std::runtime_error(Error.str());

}

float scalex = static_cast<float>(Width) / imageWidth;

float scaley = static_cast<float>(Height) / imageHeight;

fx *= scalex;

fy *= scaley;

cx *= scalex;

cy *= scaley;

imageWidth *= scalex;

imageHeight *= scaley;

imageWidth = Width;

imageHeight = Height;

}

torch::Tensor CameraIntrinsics::GetProjectionMatrix() const

{

return torch::tensor({{fx, 0.0f, cx},

{0.0f, fy, cy},

{0.0f, 0.0f, 1.0f}}, torch::kFloat32);

}

CameraTransform::CameraTransform() :

CameraTransform( {1,0,0,0, 0,1,0,0, 0,0,1,0, 0,0,0,1 })

{

}

CameraTransform::CameraTransform(const torch::Tensor& Transform) :

camToWorld ( Transform.clone().detach() )

{

// check dimensions are correct

}

CameraTransform::CameraTransform(const OpenSplat_Matrix4x4& Transform)

{

camToWorld = torch::tensor({

{Transform.m00,Transform.m01,Transform.m02,Transform.m03},

{Transform.m10,Transform.m11,Transform.m12,Transform.m13},

{Transform.m20,Transform.m21,Transform.m22,Transform.m23},

{Transform.m30,Transform.m31,Transform.m32,Transform.m33},

});

}

OpenSplat_Matrix4x4 CameraTransform::GetCamToWorldMatrix() const

{

auto& t = this->camToWorld;

OpenSplat_Matrix4x4 Out;

Out.m00 = t[0][0].item<float>();

Out.m01 = t[0][1].item<float>();

Out.m02 = t[0][2].item<float>();

Out.m03 = t[0][3].item<float>();

Out.m10 = t[1][0].item<float>();

Out.m11 = t[1][1].item<float>();

Out.m12 = t[1][2].item<float>();

Out.m13 = t[1][3].item<float>();

Out.m20 = t[2][0].item<float>();

Out.m21 = t[2][1].item<float>();

Out.m22 = t[2][2].item<float>();

Out.m23 = t[2][3].item<float>();

Out.m30 = t[3][0].item<float>();

Out.m31 = t[3][1].item<float>();

Out.m32 = t[3][2].item<float>();

Out.m33 = t[3][3].item<float>();

return Out;

}

torch::Tensor CameraTransform::GetCamToWorldRotation() const

{

torch::Tensor R = camToWorld.index({Slice(None, 3), Slice(None, 3)});

// Flip the z and y axes to align with gsplat conventions

R = torch::matmul(R, torch::diag(torch::tensor({1.0f, -1.0f, -1.0f}, R.device())));

return R;

}

torch::Tensor CameraTransform::GetWorldToCamRotation() const

{

torch::Tensor R = GetCamToWorldRotation();

// worldToCam

torch::Tensor Rinv = R.transpose(0, 1);

return Rinv;

}

torch::Tensor CameraTransform::GetWorldToCamTranslation() const

{

auto Rinv = GetWorldToCamRotation();

torch::Tensor T = GetCamToWorldTranslation();

torch::Tensor Tinv = torch::matmul(-Rinv, T);

return Tinv;

}

torch::Tensor CameraTransform::GetCamToWorldTranslation() const

{

torch::Tensor T = camToWorld.index({Slice(None, 3), Slice(3,4)});

return T;

}

void Camera::loadImageFromFilename(float downscaleFactor)

{

auto PathString = cameraImagePath.string();

std::cout << "Camera Loading Image " << PathString << "..." << std::endl;

cv::Mat cImg = imreadRGB(PathString);

loadImage( cImg, downscaleFactor );

}

void Camera::loadImage(cv::Mat& RgbPixels,float downscaleFactor)

{

auto& cImg = RgbPixels;

auto Components = cImg.channels();

if ( Components != 3 )

throw std::runtime_error("Expecting BGR image");

// Populates image and K, then updates the camera parameters

// Caution: this function has destructive behaviors

// and should be called only once

if (image.numel())

throw std::runtime_error("loadImage already called");

float rescaleF = 1.0f;

// If camera intrinsics don't match the image dimensions, rescale intrinsics to match pixels

if (cImg.rows != intrinsics.imageHeight || cImg.cols != intrinsics.imageWidth )

{

rescaleF = static_cast<float>(cImg.rows) / static_cast<float>(intrinsics.imageHeight);

intrinsics.fx *= rescaleF;

intrinsics.fy *= rescaleF;

intrinsics.cx *= rescaleF;

intrinsics.cy *= rescaleF;

// gr: should be changing imageWidth/Height in intrinsics here! - currently data is out of sync

}

// user-specified scale regardless (only scales down)

if (downscaleFactor > 1.0f)

{

float scaleFactor = 1.0f / downscaleFactor;

cv::resize(cImg, cImg, cv::Size(), scaleFactor, scaleFactor, cv::INTER_AREA);

intrinsics.fx *= scaleFactor;

intrinsics.fy *= scaleFactor;

intrinsics.cx *= scaleFactor;

intrinsics.cy *= scaleFactor;

}

// cache projection matrix

projectionMatrix = intrinsics.GetProjectionMatrix();

cv::Rect roi;

if (intrinsics.HasDistortionParameters()){

// Undistort with opencv

std::vector<float> distCoeffs = intrinsics.GetOpencvUndistortionParameters();

cv::Mat cK = floatNxNtensorToMat(projectionMatrix);

cv::Mat newK = cv::getOptimalNewCameraMatrix(cK, distCoeffs, cv::Size(cImg.cols, cImg.rows), 0, cv::Size(), &roi);

cv::Mat undistorted = cv::Mat::zeros(cImg.rows, cImg.cols, cImg.type());

cv::undistort(cImg, undistorted, cK, distCoeffs, newK);

image = imageToTensor(undistorted);

projectionMatrix = floatNxNMatToTensor(newK);

}else{

roi = cv::Rect(0, 0, cImg.cols, cImg.rows);

image = imageToTensor(cImg);

}

// Crop to ROI

image = image.index({Slice(roi.y, roi.y + roi.height), Slice(roi.x, roi.x + roi.width), Slice()});

// Update intrinsics to newly scaled parameters & image pixels

intrinsics.imageHeight = image.size(0);

intrinsics.imageWidth = image.size(1);

intrinsics.fx = projectionMatrix[0][0].item<float>();

intrinsics.fy = projectionMatrix[1][1].item<float>();

intrinsics.cx = projectionMatrix[0][2].item<float>();

intrinsics.cy = projectionMatrix[1][2].item<float>();

}

std::string Camera::getName() const

{

auto Filename = cameraImagePath.filename().string();

return Filename;

}

torch::Tensor Camera::getImage(int downscaleFactor)

{

// image isn't loaded if missing 2nd dimension (default is 0 width 1 dimension)

if ( image.dim() <= 1 )

throw std::runtime_error(std::string("No image loaded for camera ") + getName() );

if (downscaleFactor <= 1)

return image;

// torch::jit::script::Module container = torch::jit::load("gt.pt");

// return container.attr("val").toTensor();

auto ExistingIt = imagePyramids.find(downscaleFactor);

if ( ExistingIt != imagePyramids.end() )

{

return ExistingIt->second;

}

// Rescale, store and return

int NewHeight = image.size(0);

int NewWidth = image.size(1);

NewWidth /= downscaleFactor;

NewHeight /= downscaleFactor;

std::cerr << "Adding image pyramid x" << downscaleFactor << "(" << NewWidth << "x" << NewHeight << ") for " << getName() << std::endl;

cv::Mat cImg = getOpencvRgbImageStretched(NewWidth,NewHeight);

torch::Tensor t = imageToTensor(cImg);

imagePyramids[downscaleFactor] = t;

return t;

}

cv::Mat Camera::getOpencvRgbImageStretched(int Width,int Height)

{

cv::Mat SmallImage;

auto CopyToSmall = [&](const cv::Mat& Big)

{

cv::resize(Big,SmallImage, cv::Size(Width,Height), 0.0, 0.0, cv::INTER_AREA);

};

tensorToImage(image,CopyToSmall);

return SmallImage;

}

CameraIntrinsics::CameraIntrinsics(const OpenSplat_CameraIntrinsics& Intrinsics)

{

imageWidth = Intrinsics.Width;

imageHeight = Intrinsics.Height;

fx = Intrinsics.FocalWidth;

fy = Intrinsics.FocalHeight;

cx = Intrinsics.CenterX;

cy = Intrinsics.CenterY;

k1 = Intrinsics.k1;

k2 = Intrinsics.k2;

k3 = Intrinsics.k3;

p1 = Intrinsics.p1;

p2 = Intrinsics.p2;

}

bool CameraIntrinsics::HasDistortionParameters(){

return k1 != 0.0f || k2 != 0.0f || k3 != 0.0f || p1 != 0.0f || p2 != 0.0f;

}

std::vector<float> CameraIntrinsics::GetOpencvUndistortionParameters(){

std::vector<float> p = { k1, k2, p1, p2, k3, 0.0f, 0.0f, 0.0f };

return p;

}

void InputData::TransformPoints(float3 translate,float scale)

{

// apply to data

torch::Tensor translation = torch::tensor({translate.x,translate.y,translate.z});

points.xyz = (points.xyz - translation) * scale;

}

std::vector<std::string> InputData::GetCameraNames()

{

std::vector<std::string> Names;

for ( auto& Camera : cameras )

{

auto Name = Camera.getName();

Names.emplace_back( Name );

}

return Names;

}

// remove camera from the training data (typically for application to use for validation)

std::shared_ptr<Camera> InputData::PopCamera(std::string_view CameraImageName)

{

auto FindMatchingCameraIndex = [this,CameraImageName]() -> int

{

// todo: put seed into params, and use a RNG, rather than changing global (random) state

std::srand(42);

// pick a random camera index

if ( cameras.size() > 0 && CameraImageName == OpenSplat::randomValidationImageName )

{

auto Index = std::rand() % cameras.size();

return Index;

}

// find match

for ( auto i=0; i<cameras.size(); i++ )

{

auto CameraName = cameras[i].getName();

if ( CameraName != CameraImageName )

continue;

return i;

}

// helpful error listing all possibilites for the user

std::stringstream Error;

Error << "Input data has no camera matching " << CameraImageName << " (";

for ( auto& CameraName : GetCameraNames() )

Error << CameraName << ",";

Error << ")";

throw std::runtime_error(Error.str());

};

auto MatchingCameraIndex = FindMatchingCameraIndex();

auto& MatchingCamera = cameras[MatchingCameraIndex];

// make copy of camera before removing it

auto PoppedCameraCopy = std::make_shared<Camera>(MatchingCamera);

cameras.erase( cameras.begin() + MatchingCameraIndex );

return PoppedCameraCopy;

}

void InputData::saveCamerasJson(const std::string &filename)

{

json j = json::array();

for (size_t i = 0; i < cameras.size(); i++){

Camera &cam = cameras[i];

json camera = json::object();

camera["id"] = i;

camera["img_name"] = cam.getName();

camera["width"] = cam.intrinsics.imageWidth;

camera["height"] = cam.intrinsics.imageHeight;

camera["fx"] = cam.intrinsics.fx;

camera["fy"] = cam.intrinsics.fy;

torch::Tensor R = cam.camToWorld.GetCamToWorldRotation();

// gr: what is squeeze()?

torch::Tensor T = cam.camToWorld.GetCamToWorldTranslation().squeeze();

std::vector<float> position(3);

std::vector<std::vector<float>> rotation(3, std::vector<float>(3));

for (int i = 0; i < 3; i++) {

position[i] = T[i].item<float>();

for (int j = 0; j < 3; j++) {

rotation[i][j] = R[i][j].item<float>();

}

camera["position"] = position;

camera["rotation"] = rotation;

j.push_back(camera);

}

std::ofstream of(filename);

if ( !of.is_open() )

throw std::runtime_error(std::string("Failed to open file") + filename + " to write cameras meta");

of << j;

of.close();

std::cout << "Wrote " << filename << std::endl;

}

Camera& InputData::GetCamera(int CameraIndex)

{

if ( CameraIndex < 0 || CameraIndex >= cameras.size() )

{

throw OpenSplat::NoCameraException(CameraIndex,cameras.size());

}

return cameras[CameraIndex];

}

Camera& InputData::GetCamera(std::string_view CameraName)

{

for ( auto& Camera : cameras )

{

if ( Camera.getName() != CameraName )

continue;

return Camera;

}

throw std::runtime_error( std::string("No camera named ") + std::string(CameraName) );

}

void InputData::AddCamera(const Camera& newCamera)

{

if ( newCamera.getName().empty() )

throw std::runtime_error("Tring to add camera with no name");

// force names to be unique

for ( auto& Camera : cameras )

{

if ( Camera.getName() == newCamera.getName() )

throw std::runtime_error( std::string("Trying to add camera with duplicate name ") + std::string(newCamera.getName()) );

}

cameras.push_back(newCamera);

}

OpenSplat/input_data.cpp at OpenSplatLibrary · NewChromantics/OpenSplat · GitHub

Expand file tree

Breadcrumbs

input_data.cpp

Latest commit

History

Breadcrumbs

input_data.cpp

File metadata and controls