/// <summary>
/// Writes a csv file containing fish species and frame numbers.
/// </summary>
/// <param name="count_path">Path to model file.</param>
/// <param name="out_path">Path to output csv file.</param>
/// <param name="roi">Region of interest, needed for image dims.</param>
/// <param name="detections">Detections for each frame.</param>
/// <param name="scores">Cover and species scores for each detection.</param>
static void WriteCounts(
string count_path,
string out_path,
Rect roi,
VectorVectorDetection detections,
VectorVectorClassification scores)
{
// Create and initialize keyframe finder.
KeyframeFinder finder = new KeyframeFinder();
ErrorCode status = finder.Init(count_path, roi[2], roi[3]);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to initialize keyframe finder!");
}
// Process keyframe finder.
VectorInt keyframes = new VectorInt();
status = finder.Process(scores, detections, keyframes);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to process keyframe finder!");
}
// Write the keyframes out.
using (var csv = new System.IO.StreamWriter(out_path)) {
csv.WriteLine("id,frame,species_index");
int id = 0;
foreach (var i in keyframes)
{
Classification c = scores[i][0];
float max_score = 0.0F;
int species_index = 0;
for (int j = 0; j < c.species.Count; ++j)
{
if (c.species[j] > max_score)
{
max_score = c.species[j];
species_index = j;
}
}
var line = string.Format("{0},{1},{2}", id, i, species_index);
csv.WriteLine(line);
id++;
}
}
}
/// <summary>
/// Writes a new video with bounding boxes around detections.
/// </summary>
/// <param name="vid_path">Path to the original video.</param>
/// <param name="out_path">Path to the output video.</param>
/// <param name="roi">Region of interest output from FindVideoRoi.</param>
/// <param name="endpoints">Transform output from FindVideoRoi.</param>
/// <param name="detections">Detections for each frame.</param>
/// <param name="scores">Cover and species scores for each detection.</param>
static void WriteVideo(
string vid_path,
string out_path,
Rect roi,
PointPair endpoints,
VectorVectorDetection detections,
VectorVectorClassification scores)
{
// Initialize the video reader.
VideoReader reader = new VideoReader();
ErrorCode status = reader.Init(vid_path);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to read video!");
}
// Initialize the video writer.
VideoWriter writer = new VideoWriter();
status = writer.Init(
out_path,
reader.FrameRate(),
Codec.kWmv2,
new PairIntInt(reader.Width(), reader.Height()));
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to write video!");
}
// Iterate through frames.
for (int i = 0; i < detections.Count; ++i)
{
Image frame = new Image();
status = reader.GetFrame(frame);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Error retrieving video frame!");
}
Color blue = new Color(); blue[0] = 255; blue[1] = 0; blue[2] = 0;
Color red = new Color(); red[0] = 0; red[1] = 0; red[2] = 255;
Color green = new Color(); green[0] = 0; green[1] = 255; green[2] = 0;
frame.DrawRect(roi, blue, 1, endpoints);
for (int j = 0; j < detections[i].Count; ++j)
{
Color det_color = red;
double clear = scores[i][j].cover[2];
double hand = scores[i][j].cover[1];
if (j == 0)
{
if (clear > hand)
{
det_color = green;
}
else
{
det_color = red;
}
}
frame.DrawRect(detections[i][j].location, det_color, 2, endpoints, roi);
}
status = writer.AddFrame(frame);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Error adding frame to video!");
}
}
}
/// <summary>
/// Finds and classifies detections for all frames in a video.
/// </summary>
/// <param name="detect_path">Path to detect model file.</param>
/// <param name="classify_path">Path to classify model file.</param>
/// <param name="roi">Region of interest output from FindVideoRoi.</param>
/// <param name="endpoints">Ruler endpoints output from FindVideoRoi.</param>
/// <param name="detections">Detections for each frame.</param>
/// <param name="scores">Cover and species scores for each detection.</param>
static void DetectAndClassify(
string detect_path,
string classify_path,
string vid_path,
Rect roi,
PointPair endpoints,
out VectorVectorDetection detections,
out VectorVectorClassification scores)
{
// Determined by experimentation with GPU having 8GB memory.
const int kMaxImg = 32;
// Initialize the outputs.
detections = new VectorVectorDetection();
scores = new VectorVectorClassification();
// Create and initialize the detector.
Detector detector = new Detector();
ErrorCode status = detector.Init(detect_path, 0.5);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to initialize detector!");
}
// Create and initialize the classifier.
Classifier classifier = new Classifier();
status = classifier.Init(classify_path, 0.5);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to initialize classifier!");
}
// Initialize the video reader.
VideoReader reader = new VideoReader();
status = reader.Init(vid_path);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to open video!");
}
// Iterate through frames.
bool vid_end = false;
while (true)
{
// Find detections.
VectorVectorDetection dets = new VectorVectorDetection();
VectorImage imgs = new VectorImage();
for (int i = 0; i < kMaxImg; ++i)
{
Image img = new Image();
status = reader.GetFrame(img);
if (status != ErrorCode.kSuccess)
{
vid_end = true;
break;
}
img = openem.Rectify(img, endpoints);
img = openem.Crop(img, roi);
imgs.Add(img);
status = detector.AddImage(img);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to add frame to detector!");
}
}
status = detector.Process(dets);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to process detector!");
}
for (int i = 0; i < dets.Count; ++i)
{
}
detections.AddRange(dets);
for (int i = 0; i < detections.Count; ++i)
{
}
// Classify detections.
for (int i = 0; i < dets.Count; ++i)
{
VectorClassification score_batch = new VectorClassification();
for (int j = 0; j < dets[i].Count; ++j)
{
Image det_img = openem.GetDetImage(imgs[i], dets[i][j].location);
status = classifier.AddImage(det_img);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to add frame to classifier!");
}
}
status = classifier.Process(score_batch);
if (status != ErrorCode.kSuccess)
{
throw new Exception("Failed to process classifier!");
}
scores.Add(score_batch);
}
if (vid_end)
{
break;
}
}
}
