Fix issues with seeking in some H.264 videos on native & web

Cargo.lock

@@ -5823,9 +5823,9 @@ dependencies = [
 
 [[package]]
 name = "re_mp4"
-version = "0.2.1"
+version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "cb588af9f28ae35e59681c71c0d6379884e1209a074ffb1b3b18291abac2585a"
+checksum = "751650322999417b64a5a89b034b4e34e4596826e5dfee9327856db77ca511e3"
 dependencies = [
  "byteorder",
  "bytes",

Cargo.toml

@@ -206,7 +206,7 @@ memory-stats = "1.1"
 mimalloc = "0.1.43"
 mime_guess2 = "2.0" # infer MIME type by file extension, and map mime to file extension
 mint = "0.5.9"
-re_mp4 = "0.2.1"
+re_mp4 = "0.3.0"
 natord = "1.0.9"
 ndarray = "0.16"
 ndarray-rand = "0.15"

crates/store/re_video/src/decode/webcodecs.rs

@@ -34,7 +34,6 @@ impl std::ops::Deref for WebVideoFrame {
 pub struct WebVideoDecoder {
     video_config: Config,
     timescale: Timescale,
-    minimum_presentation_timestamp: Time,
     decoder: web_sys::VideoDecoder,
     hw_acceleration: DecodeHardwareAcceleration,
     on_output: Arc<OutputCallback>,
@@ -108,16 +107,11 @@ impl WebVideoDecoder {
         on_output: impl Fn(Result<Frame>) + Send + Sync + 'static,
     ) -> Result<Self, Error> {
         let on_output = Arc::new(on_output);
-        let decoder = init_video_decoder(
-            on_output.clone(),
-            video.timescale,
-            video.samples_statistics.minimum_presentation_timestamp,
-        )?;
+        let decoder = init_video_decoder(on_output.clone(), video.timescale)?;
 
         Ok(Self {
             video_config: video.config.clone(),
             timescale: video.timescale,
-            minimum_presentation_timestamp: video.samples_statistics.minimum_presentation_timestamp,
             decoder,
             hw_acceleration,
             on_output,
@@ -138,7 +132,7 @@ impl AsyncDecoder for WebVideoDecoder {
             &data,
             video_chunk
                 .presentation_timestamp
-                .into_micros_since_start(self.timescale, self.minimum_presentation_timestamp),
+                .into_micros(self.timescale),
             type_,
         );
 
@@ -162,11 +156,7 @@ impl AsyncDecoder for WebVideoDecoder {
             // At least on Firefox, it can happen that reset on a previous error fails.
             // In that case, start over completely and try again!
             re_log::debug!("Video decoder reset failed, recreating decoder.");
-            self.decoder = init_video_decoder(
-                self.on_output.clone(),
-                self.timescale,
-                self.minimum_presentation_timestamp,
-            )?;
+            self.decoder = init_video_decoder(self.on_output.clone(), self.timescale)?;
         };
 
         self.decoder
@@ -183,23 +173,15 @@ impl AsyncDecoder for WebVideoDecoder {
 fn init_video_decoder(
     on_output_callback: Arc<OutputCallback>,
     timescale: Timescale,
-    minimum_presentation_timestamp: Time,
 ) -> Result<web_sys::VideoDecoder, Error> {
     let on_output = {
         let on_output = on_output_callback.clone();
         Closure::wrap(Box::new(move |frame: web_sys::VideoFrame| {
             // We assume that the timestamp returned by the decoder is in time since start,
             // and does not represent demuxed "raw" presentation timestamps.
-            let presentation_timestamp = Time::from_micros_since_start(
-                frame.timestamp().unwrap_or(0.0),
-                timescale,
-                minimum_presentation_timestamp,
-            );
-            let duration = Time::from_micros_since_start(
-                frame.duration().unwrap_or(0.0),
-                timescale,
-                minimum_presentation_timestamp,
-            );
+            let presentation_timestamp =
+                Time::from_micros(frame.timestamp().unwrap_or(0.0), timescale);
+            let duration = Time::from_micros(frame.duration().unwrap_or(0.0), timescale);
 
             on_output(Ok(Frame {
                 content: WebVideoFrame(frame),

crates/store/re_video/src/demux/mod.rs

@@ -75,19 +75,6 @@ pub struct VideoData {
 /// Meta informationa about the video samples.
 #[derive(Clone, Debug)]
 pub struct SamplesStatistics {
-    /// The smallest presentation timestamp observed in this video.
-    ///
-    /// This is typically 0, but in the presence of B-frames, it may be non-zero.
-    /// In fact, many formats don't require this to be zero, but video players typically
-    /// normalize the shown time to start at zero.
-    /// Note that timestamps in the [`Sample`]s are *not* automatically adjusted with this value.
-    // This is roughly equivalent to FFmpeg's internal `min_corrected_pts`
-    // https://github.com/FFmpeg/FFmpeg/blob/4047b887fc44b110bccb1da09bcb79d6e454b88b/libavformat/isom.h#L202
-    // (unlike us, this handles a bunch more edge cases but it fulfills the same role)
-    // To learn more about this I recommend reading the patch that introduced this in FFmpeg:
-    // https://patchwork.ffmpeg.org/project/ffmpeg/patch/[email protected]/#12592
-    pub minimum_presentation_timestamp: Time,
-
     /// Whether all decode timestamps are equal to presentation timestamps.
     ///
     /// If true, the video typically has no B-frames as those require frame reordering.
@@ -103,11 +90,6 @@ impl SamplesStatistics {
     pub fn new(samples: &[Sample]) -> Self {
         re_tracing::profile_function!();
 
-        let minimum_presentation_timestamp = samples
-            .iter()
-            .map(|s| s.presentation_timestamp)
-            .min()
-            .unwrap_or_default();
         let dts_always_equal_pts = samples
             .iter()
             .all(|s| s.decode_timestamp == s.presentation_timestamp);
@@ -128,7 +110,6 @@ impl SamplesStatistics {
         });
 
         Self {
-            minimum_presentation_timestamp,
             dts_always_equal_pts,
             has_sample_highest_pts_so_far,
         }
@@ -301,8 +282,6 @@ impl VideoData {
     /// Determines the video timestamps of all frames inside a video, returning raw time values.
     ///
     /// Returned timestamps are in nanoseconds since start and are guaranteed to be monotonically increasing.
-    /// These are *not* necessarily the same as the presentation timestamps, as the returned timestamps are
-    /// normalized respect to the start of the video, see [`SamplesStatistics::minimum_presentation_timestamp`].
     pub fn frame_timestamps_ns(&self) -> impl Iterator<Item = i64> + '_ {
         // Segments are guaranteed to be sorted among each other, but within a segment,
         // presentation timestamps may not be sorted since this is sorted by decode timestamps.
@@ -311,12 +290,7 @@ impl VideoData {
                 .iter()
                 .map(|sample| sample.presentation_timestamp)
                 .sorted()
-                .map(|pts| {
-                    pts.into_nanos_since_start(
-                        self.timescale,
-                        self.samples_statistics.minimum_presentation_timestamp,
-                    )
-                })
+                .map(|pts| pts.into_nanos(self.timescale))
         })
     }
 
@@ -638,16 +612,16 @@ mod tests {
     fn test_latest_sample_index_at_presentation_timestamp() {
         // This is a snippet of real world data!
         let pts = [
-            512, 1536, 1024, 768, 1280, 2560, 2048, 1792, 2304, 3584, 3072, 2816, 3328, 4608, 4096,
-            3840, 4352, 5376, 4864, 5120, 6400, 5888, 5632, 6144, 7424, 6912, 6656, 7168, 8448,
-            7936, 7680, 8192, 9472, 8960, 8704, 9216, 10496, 9984, 9728, 10240, 11520, 11008,
-            10752, 11264, 12544, 12032, 11776, 12288, 13568, 13056,
+            0, 1024, 512, 256, 768, 2048, 1536, 1280, 1792, 3072, 2560, 2304, 2816, 4096, 3584,
+            3328, 3840, 4864, 4352, 4608, 5888, 5376, 5120, 5632, 6912, 6400, 6144, 6656, 7936,
+            7424, 7168, 7680, 8960, 8448, 8192, 8704, 9984, 9472, 9216, 9728, 11008, 10496, 10240,
+            10752, 12032, 11520, 11264, 11776, 13056, 12544,
         ];
         let dts = [
-            0, 256, 512, 768, 1024, 1280, 1536, 1792, 2048, 2304, 2560, 2816, 3072, 3328, 3584,
-            3840, 4096, 4352, 4608, 4864, 5120, 5376, 5632, 5888, 6144, 6400, 6656, 6912, 7168,
-            7424, 7680, 7936, 8192, 8448, 8704, 8960, 9216, 9472, 9728, 9984, 10240, 10496, 10752,
-            11008, 11264, 11520, 11776, 12032, 12288, 12544,
+            -512, -256, 0, 256, 512, 768, 1024, 1280, 1536, 1792, 2048, 2304, 2560, 2816, 3072,
+            3328, 3584, 3840, 4096, 4352, 4608, 4864, 5120, 5376, 5632, 5888, 6144, 6400, 6656,
+            6912, 7168, 7424, 7680, 7936, 8192, 8448, 8704, 8960, 9216, 9472, 9728, 9984, 10240,
+            10496, 10752, 11008, 11264, 11520, 11776, 12032,
         ];
 
         // Checking our basic assumptions about this data:
@@ -671,7 +645,6 @@ mod tests {
             .collect::<Vec<_>>();
 
         let sample_statistics = SamplesStatistics::new(&samples);
-        assert_eq!(sample_statistics.minimum_presentation_timestamp, Time(512));
         assert!(!sample_statistics.dts_always_equal_pts);
 
         // Test queries on the samples.
@@ -704,30 +677,30 @@ mod tests {
         // A few hardcoded cases - both for illustrative purposes and to make sure the generic tests above are correct.
 
         // Querying before the first sample.
-        assert_eq!(None, query_pts(Time(0)));
-        assert_eq!(None, query_pts(Time(123)));
+        assert_eq!(None, query_pts(Time(-1)));
+        assert_eq!(None, query_pts(Time(-123)));
 
         // Querying for the first sample
-        assert_eq!(Some(0), query_pts(Time(512)));
-        assert_eq!(Some(0), query_pts(Time(513)));
-        assert_eq!(Some(0), query_pts(Time(600)));
-        assert_eq!(Some(0), query_pts(Time(767)));
+        assert_eq!(Some(0), query_pts(Time(0)));
+        assert_eq!(Some(0), query_pts(Time(1)));
+        assert_eq!(Some(0), query_pts(Time(88)));
+        assert_eq!(Some(0), query_pts(Time(255)));
 
         // The next sample is a jump in index!
-        assert_eq!(Some(3), query_pts(Time(768)));
-        assert_eq!(Some(3), query_pts(Time(769)));
-        assert_eq!(Some(3), query_pts(Time(800)));
-        assert_eq!(Some(3), query_pts(Time(1023)));
+        assert_eq!(Some(3), query_pts(Time(256)));
+        assert_eq!(Some(3), query_pts(Time(257)));
+        assert_eq!(Some(3), query_pts(Time(400)));
+        assert_eq!(Some(3), query_pts(Time(511)));
 
         // And the one after that should jump back again.
-        assert_eq!(Some(2), query_pts(Time(1024)));
-        assert_eq!(Some(2), query_pts(Time(1025)));
-        assert_eq!(Some(2), query_pts(Time(1100)));
-        assert_eq!(Some(2), query_pts(Time(1279)));
+        assert_eq!(Some(2), query_pts(Time(512)));
+        assert_eq!(Some(2), query_pts(Time(513)));
+        assert_eq!(Some(2), query_pts(Time(600)));
+        assert_eq!(Some(2), query_pts(Time(767)));
 
         // And another one!
-        assert_eq!(Some(4), query_pts(Time(1280)));
-        assert_eq!(Some(4), query_pts(Time(1281)));
+        assert_eq!(Some(4), query_pts(Time(768)));
+        assert_eq!(Some(4), query_pts(Time(1023)));
 
         // Test way outside of the range.
         // (this is not the last element in the list since that one doesn't have the highest PTS)

crates/store/re_video/src/demux/mp4.rs

@@ -56,8 +56,8 @@ impl VideoData {
                     gop_sample_start_index = samples.len();
                 }
 
-                let decode_timestamp = Time::new(sample.decode_timestamp as i64);
-                let presentation_timestamp = Time::new(sample.composition_timestamp as i64);
+                let decode_timestamp = Time::new(sample.decode_timestamp);
+                let presentation_timestamp = Time::new(sample.composition_timestamp);
                 let duration = Time::new(sample.duration as i64);
 
                 let byte_offset = sample.offset as u32;
@@ -75,6 +75,24 @@ impl VideoData {
             }
         }
 
+        // Generate data for `test_latest_sample_index_at_presentation_timestamp` test.
+        // re_log::info!(
+        //     "pts: {:?}",
+        //     samples
+        //         .iter()
+        //         .take(50)
+        //         .map(|s| s.presentation_timestamp.0)
+        //         .collect::<Vec<_>>()
+        // );
+        // re_log::info!(
+        //     "dts: {:?}",
+        //     samples
+        //         .iter()
+        //         .take(50)
+        //         .map(|s| s.decode_timestamp.0)
+        //         .collect::<Vec<_>>()
+        // );
+
         if !samples.is_empty() {
             let start = samples[gop_sample_start_index].decode_timestamp;
             let sample_range = gop_sample_start_index as u32..samples.len() as u32;

crates/store/re_video/src/time.rs

@@ -28,67 +28,50 @@ impl Time {
         Self(v)
     }
 
-    /// `time_base` specifies the
     #[inline]
-    pub fn from_secs_since_start(
-        secs_since_start: f64,
-        timescale: Timescale,
-        start_time: Self,
-    ) -> Self {
-        Self((secs_since_start * timescale.0 as f64).round() as i64 + start_time.0)
+    pub fn from_secs(secs_since_start: f64, timescale: Timescale) -> Self {
+        Self((secs_since_start * timescale.0 as f64).round() as i64)
     }
 
     #[inline]
-    pub fn from_millis_since_start(
-        millis_since_start: f64,
-        timescale: Timescale,
-        start_time: Self,
-    ) -> Self {
-        Self::from_secs_since_start(millis_since_start / 1e3, timescale, start_time)
+    pub fn from_millis(millis_since_start: f64, timescale: Timescale) -> Self {
+        Self::from_secs(millis_since_start / 1e3, timescale)
     }
 
     #[inline]
-    pub fn from_micros_since_start(
-        micros_since_start: f64,
-        timescale: Timescale,
-        start_time: Self,
-    ) -> Self {
-        Self::from_secs_since_start(micros_since_start / 1e6, timescale, start_time)
+    pub fn from_micros(micros_since_start: f64, timescale: Timescale) -> Self {
+        Self::from_secs(micros_since_start / 1e6, timescale)
     }
 
     #[inline]
-    pub fn from_nanos_since_start(
-        nanos_since_start: i64,
-        timescale: Timescale,
-        start_time: Self,
-    ) -> Self {
-        Self::from_secs_since_start(nanos_since_start as f64 / 1e9, timescale, start_time)
+    pub fn from_nanos(nanos_since_start: i64, timescale: Timescale) -> Self {
+        Self::from_secs(nanos_since_start as f64 / 1e9, timescale)
     }
 
     /// Convert to a duration
     #[inline]
     pub fn duration(self, timescale: Timescale) -> std::time::Duration {
-        std::time::Duration::from_nanos(self.into_nanos_since_start(timescale, Self(0)) as _)
+        std::time::Duration::from_nanos(self.into_nanos(timescale) as _)
     }
 
     #[inline]
-    pub fn into_secs_since_start(self, timescale: Timescale, start_time: Self) -> f64 {
-        (self.0 - start_time.0) as f64 / timescale.0 as f64
+    pub fn into_secs(self, timescale: Timescale) -> f64 {
+        self.0 as f64 / timescale.0 as f64
     }
 
     #[inline]
-    pub fn into_millis_since_start(self, timescale: Timescale, start_time: Self) -> f64 {
-        self.into_secs_since_start(timescale, start_time) * 1e3
+    pub fn into_millis(self, timescale: Timescale) -> f64 {
+        self.into_secs(timescale) * 1e3
     }
 
     #[inline]
-    pub fn into_micros_since_start(self, timescale: Timescale, start_time: Self) -> f64 {
-        self.into_secs_since_start(timescale, start_time) * 1e6
+    pub fn into_micros(self, timescale: Timescale) -> f64 {
+        self.into_secs(timescale) * 1e6
     }
 
     #[inline]
-    pub fn into_nanos_since_start(self, timescale: Timescale, start_time: Self) -> i64 {
-        (self.into_secs_since_start(timescale, start_time) * 1e9).round() as i64
+    pub fn into_nanos(self, timescale: Timescale) -> i64 {
+        (self.into_secs(timescale) * 1e9).round() as i64
     }
 }