Added face detection

src/state.rs

@@ -46,6 +46,7 @@ impl TelloState {
 
  let str_num = String::from_iter(num);
 
+ // Very ugly parsing solution, pretty sure there's something better out there
  match i {
  0 => tellostate.mid = str_num.parse().unwrap(),
  (1..=3) => tellostate.xyz[i-1] = str_num.parse().unwrap(),

src/video.rs

@@ -2,8 +2,10 @@
 use opencv as cv;
 use cv::{
  highgui,
- objdetect::CascadeClassifier,
- types::VectorOfRect,
+ objdetect::{CascadeClassifier, self},
+ types::VectorOfRect, prelude::CascadeClassifierTrait,
+ core::Size,
+ imgproc::rectangle,
 };
 
 mod decoding;
@@ -24,6 +26,14 @@ pub fn spawn_video_thread(vrx: std::sync::mpsc::Receiver<crate::ThreadMsg>) -> s
 
  let mut decoder = decoding::Decoder::new().unwrap();
 
+ let mut face_detector = match CascadeClassifier::new("./cascades/haarcascade_frontalface_alt.xml") {
+ Ok(cc) => cc,
+ Err(e) => panic!("Failed to create face detector: {:?}", e),
+ };
+
+ let mut faces = VectorOfRect::new();
+ let mut find_face_in = 3;
+
  loop {
  let mut video_buffer = [0; 2048];
  let msg_len = match video_socket.recv(&mut video_buffer) {
@@ -32,23 +42,18 @@ pub fn spawn_video_thread(vrx: std::sync::mpsc::Receiver<crate::ThreadMsg>) -> s
  };
 
  frame_packet.extend_from_slice(&video_buffer[..msg_len]);
-
+ 
  match decoding::check_for_valid_packet(&frame_packet) {
  Some(frame_borders) => {
  match decoding::h264_decode(&frame_packet[(frame_borders.0)..(frame_borders.1)], &mut decoder) {
  Ok(mut result) => {
  frame_packet = frame_packet[(frame_borders.1)..(frame_packet.len())].to_vec();
 
- let bgra_frame = match conversions::vec_to_bgra_mat(&mut result.0, result.1) {
+ let mut bgra_frame = match conversions::vec_to_bgra_mat(&mut result.0, result.1) {
  Ok(m) => m,
  Err(()) => continue,
  };
 
- match highgui::imshow("tello", &bgra_frame) {
- Ok(_) => {},
- Err(e) => println!("Error: {:?}", e),
- };
-
  let gray_buf = match conversions::mat_to_grayscale(&bgra_frame, result.1) {
  Ok(m) => m,
  Err(_) => continue,
@@ -59,6 +64,46 @@ pub fn spawn_video_thread(vrx: std::sync::mpsc::Receiver<crate::ThreadMsg>) -> s
  Err(e) => println!("Error: {:?}", e),
  };
 
+ find_face_in -= 1;
+
+ // We need to limit how much the classifier is ran, for performance
+ // reasons.
+ if find_face_in == 0 {
+ faces = VectorOfRect::new();
+ match face_detector.detect_multi_scale(
+ &gray_buf,
+ &mut faces,
+ 1.1,
+ 2,
+ objdetect::CASCADE_SCALE_IMAGE,
+ Size::new(30, 30),
+ Size::new(0, 0)
+ ) {
+ Ok(_) => {},
+ Err(e) => println!("Error with face detector: {}", e),
+ };
+
+ find_face_in = 3;
+ }
+
+ // We can still display the last frames rectangle though, since it's
+ // quite impossible to notice changes in that little amount of time.
+ for face in &faces {
+ rectangle(
+ &mut bgra_frame,
+ face,
+ cv::core::Scalar::new(255.0, 0.0, 0.0, 0.0),
+ 2,
+ cv::imgproc::LINE_8,
+ 0
+ ).unwrap();
+ }
+
+ match highgui::imshow("tello", &bgra_frame) {
+ Ok(_) => {},
+ Err(e) => println!("Error: {:?}", e),
+ };
+
  highgui::poll_key();
 
  match vrx.try_recv() {